Home

Guide, Axiom Genotyping Solution Analysis

Contents

1. Sample Table tab 208 g 3 enin SNP Summary Table tab Show ProPMedele Z Show Posterior Models SNP Summary Table Sample Table 2 50 5 00 Default View Minor Allele Probe Set ID SNP Call Rate SNP ZAA SNP XAB SNP BB Frequency SNP H W pValue dbs Summary AX 11089485 97 78 3 333333 37 77778 56 66667 0 2272727 0 3482017 AX 11089487 100 24 44444 57 77778 17 77778 0 4666667 0 1273415 AX 11089483 100 94 44444 5 555555 0 0 02777778 0 7863507 AX 11089490 13 0 3611111 0 9040329 Status bar Rows 45191 Cols 12 Selected Rows 1 The SNP Cluster Graph function has several drop down menus will allow the user to change the genotype color the posterior and prior distribution colors and the shapes used for the samples Figure 7 13 shows a cluster plot where the AA cluster is red the AB cluster is blue the BB cluster is green and female samples are plotted as circles while male samples are plotted as squares For more information on adjusting color and shape see the Genotyping Console User Guide 4 2 P N 702982 chapter 9 Using the SNP Cluster Graph Display a Particular SNP To display a particular SNP click the corresponding row in the SNP Summary Table The cluster graph will update to display the data for the SNP Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software Select a Single Sample To select a single sample click the data point in the SNP cluste
2. Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 66 Figure 8 1 An example of the output file from Ps Metrics AX 89778338 99 6479 5 9653 NA_ 0 38863 269981 amp 9 2 3194 sf oo 1 O Ax 89778339 100NA na NA 167249 o i o o o o JAX 89778340 99 6479 5 6528 NA 013211 139684 55 2 o ss 28 i o o AX 89778341 96 4789 46887 NA___ 0 13635 118060 ej 2 of esj 2oe 10 ol O ojojojojo gt AX 89778346 97 1831 S0842 NA 0 19844 194264 s 2 of so a ao o Ax 89778348 100 amp 750 NA o65794 230562 1 2 j 3 oA oA o jx 89778349 98 5915 4sO53 NA o22793 185183 uoj af of us 161 aA o cec O o oljolijo The first 13 rows of the output file from Ps Metrics metrics txt opened with Excel Best Practices Step 8B Run Ps Classification Once the Ps Metrics function has been run and the SNP QC metrics generated and output to metrics txt the SNPs can be classified using Ps Classification Ps Classification has three required arguments and 15 optional arguments The three required arguments are 1 the name and location of the output metrics file from Ps Metrics 2 the location of the preferred output directory and 3 the species or genome type human non human diploid or polyploid A 4th argument ps2snpFile is needed for array
3. affymetrix A nal V sr Cttide Axiom Genotyping Solution Data Analysis Guide For Research Use Only Not for use in diagnostic procedures Trademarks Affymetrix Axiom Command Console CytoScan DMET GeneAtlas GeneChip GeneChip compatible GeneTitan Genotyping Console myDesign NetAffx OncoScan Powered by Affymetrix PrimeView Procarta and QuantiGene are trademarks or registered trademarks of Affymetrix Inc All other trademarks are the property of their respective owners Limited License Subject to the Affymetrix terms and conditions that govern your use of Affymetrix products Affymetrix grants you a non exclusive non transferable non sublicensable license to use this Affymetrix product only in accordance with the manual and written instructions provided by Affymetrix You understand and agree that except as expressly set forth in the Affymetrix terms and conditions no right or license to any patent or other intellectual property owned or licensable by Affymetrix is conveyed or implied by this Affymetrix product In particular no right or license is conveyed or implied to use this Affymetrix product in combination with a product not provided licensed or specifically recommended by Affymetrix for such use Patents Axiom myDesign Genotyping Arrays may be covered by one or more of the following patents U S Patent Nos 6 307 042 6 706 875 7 332 273 7 790 389 8 114 584 8 273 3
4. Alternatively samples can be grouped into batches outside GTC and only the CEL files of interest should be imported into a dataset during GTC data import Steps 2 and 3 Generate DQC Values and QC the Samples Based on DQC in GTC To perform a sample quality check in GTC using DQC values import the CEL files of interest into a dataset created for the specific Axiom array For instructions on creating a dataset in GTC and for general instructions on working with GTC consult the Genotyping Console User Guide P N 702982 GTC will calculate the quality control metrics at the time of import or at any other time by performing any one of the following operations Select the QC icon QC from the tool bar menu Figure 7 1 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 42 g Workspace Axiom_GW_LAT Data Sets S Axiom_GW_LAT Axiom KP UCSF LAT Ej Sample Attributes 3 Intensity Data Select a user intensity data group Ei P Select the Perform Intensity QC icon from the GTC tool bar menu In Bounds Out of Bounds E Genotype Results Copy Number LOH Results Copy Number Variation Results Reports E SNP Lists Select the Intensity Data option from the Workspace drop down menu and then select Perform QC Figure 7 2 Results Reports Copy Number LOH Results Copy Number Variation Results SNP Lists Perform Genotyping Copy Number LOH Analysis Perfo
5. GTC contains a function for plotting SNP cluster graphs What is a SNP Cluster Plot for AxiomGTI1 Genotypes that produces plots that are similar to the output from the R function Ps Visualization However the SNP Cluster Graph function in GTC 4 2 has more functionality than the Ps Visualization function We strongly suggest using the GTC 4 2 SNP Cluster Graph function with the output generated by Ps Metrics and Ps Classification For a more detailed introduction to the SNP Cluster Graph function see the Genotyping Console User Guide 4 2 P N 702982 chapter 9 Using the SNP Cluster Graph Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 53 SNP Cluster Graph allows the user to adjust the shape and color of the samples and posterior and prior distributions and users can select and change the calls of samples through the plotted cluster graph See Evaluate SNP Cluster Plots on page 23 for interpretation of cluster graphs In order to visualize probe sets listed in the category files from APT Ps_Classification PolyHighResolution txt NoMinorHom txt Hemizygous txt MonoHighResolution txt CallRateBelowThreshold txt Other txt and OTV txt described in Step SB Classify SNPs Using OC Metrics on page 19 these files must be imported into GTC as a SNP list The GTC SNP list file must be a text file with a txt or tsv extension and contain a column labeled Probe Set ID The category files from APT Ps
6. to be written by the user The example scripts below are directly usable by Linux users For readability the Linux commands are broken into several lines with the backslash character The backslash character is not recognized by the Windows OS The APT commands can also be executed on a Windows computer To execute the commands scripts in Windows 1 Remove the backslashes and put the given command on one line 2 Change the forward slash to a backslash when the input is a directory path 3 Enter the command in the Windows command prompt window Best Practices Step 1 Group Samples into Batches In preparation for step 2 of the best practice analysis workflow with APT the Generate the Sample DQC values step CEL files corresponding to each batch must be collected into a file we will refer to the files within each array batch as the cel_list with the full path to each CEL file in each row and with a header line cel files We will refer to this list as cel listl txt Below is a useful Linux one liner for making cel lists echo cel files Ms 1 DIRECTORY CONTAINING CEL FILES gt CEL gt lt OUTDIR gt cel_list1 txt Best Practices Step 2 Generate the Sample DQC Values Using APT DQC values are produced by the program apt geno qc Below is an example script for a Linux command line shell In this and other example scripts with APT programs we assume that the analysis
7. 2 00 eee 66 Visualize SNP Cluster Plots with SNPolisher Ps Visualization Function 67 References 2 0 esser 71 Introduction to Axiom Data Analysis About this Guide Purpose This guide provides information and instructions for analyzing Axiom genotyping array data It includes the use of Affymetrix Genotyping Console Software GTC or Affymetrix Power Tools APT and SNPolisher R package to perform quality control analysis QC for samples plates and SNP filtering prior to downstream analysis and advanced genotyping methods While this guide contains specific information tailored to analyzing Axiom genotyping array data most principles can be applied to all Affymetrix genotyping array data with the QC metrics being array specific e g contrast QC for Genome Wide SNP 6 0 Arrays vs dish QC for Axiom arrays Prerequisites Support This guide is intended for scientists technicians and bioinformaticians who need to analyze Axiom genotyping array data This guide uses conventions and terminology that assume a working knowledge of bioinformatics microarrays association studies quality control and data normalization analysis Users should contact their local Affymetrix Field Application Support or send email to Support affymetrix com Analysis Software Three analysis software systems are used for Axiom analysis and described in this document 1 Affymetrix Genotyping Console GTC version 4
8. 2 and above 2 Affymetrix Power Tools APT version 1 16 1 and above 3 the SNPolisher R package version 1 5 0 and above The workflow utilizing these software systems is shown in the section Execute the Required Steps of the Workflow on page 14 GTC is a software package that provides a graphical user interface for most of the algorithms contained in APT and is designed to streamline whole genome genotyping analysis and quality control Genotyping Console Software information GTC includes quality control and visualization tools to easily identify and segregate sample outliers including a cluster graph visualization tool that enables a detailed look at the performance of SNPs of interest GTC 4 2 and above provides the capability to manually edit genotypes APT is a set of cross platform command line programs that implement algorithms for analyzing and working with Affymetrix arrays Affymetrix Power Tools information APT programs are intended for expert users who prefer programs that can be utilized in scripting environments and are sophisticated enough to handle the complexity of extra features and functionality For more information on the setup and operation of these tools please refer to the GTC software user manual and the APT help SNPolisher R functions provide SNP quality control and classification visualization tools and advanced genotyping methods Many of the SNPolisher functions are available in APT or GTC Usage of SNP
9. Call Rate and HetSO tests Evaluate SNP Cluster Plots Visualization and understanding of SNP cluster plots introduced in What is a SNP Cluster Plot for AxiomGTI Genotypes on page 10 is a key component the Best Practices workflow Users should view a small number 200 of cluster plots of randomly selected SNPs from each of the Ps Classification function categories Figure 3 3 in order to check that SNPs have the expected cluster plot patterns for the category SNPs with mis clustered multi clustered and or poorly resolved clusters plots should be sorted into CallRateBelowThreshold or Other classes SNPs in the default recommended categories Table 3 3 should have clusters are that are reasonably separated from one another have no visible batch effects or other cluster anomalies and should not appear to be of the OTV type SNPs in the OTV class should have a four cluster OTV pattern Functions for creating SNP cluster plots are provided by two Axiom software systems 1 GTC via the SNP Cluster Graph function and 2 the SNPolisher package via the Ps Visualization function Instructions for the Cluster Graph and Ps Visualization function usages are provided in Chapter 7 and Chapter 8 respectively Cluster plots in this section were produced by the Ps Visualization function Chapter 3 Best Practices Genotyping Analysis Workflow 24 Well clustered vs Mis clustered SNP Cluster Plot Patterns Figure 3 5 shows an example of a probe
10. Genotypes produced by this step are only for the purpose of Sample QC and are not intended for downstream analysis Step 5 QC the Samples Based on QC Call Rate Samples with a QC call rate value less than the default threshold of 9796 should be excluded from step 7 genotyping Such samples should be either reprocessed in the laboratory or dropped from the study Steps 3 and 5 are the sample QC tests developed for Axiom arrays and are the minimum requirements of the Best Practices workflow See Additional Sample QC on page 30 for additional Axiom methods and general methods used in the field to detect outlier and problem samples Step 6 QC the Plates For Axiom genotyping projects samples are processed together on a 96 or 384 array plate In step 6 basic plate QC metrics are computed and all samples on plates with non passing QC metrics should be excluded from the final genotyping run which will be executed in step 7 of the workflow The specification for a non passing plate is when the average QC call rate of passing samples passing steps 2 5 is less than 98 5 The reason for including a plate QC test in the Best Practices workflow is that plates whose sample intensities systematically differ from other plates for some probe sets may contribute to mis clustering events described in Evaluate SNP Cluster Plots on page 23 whether processed separately or processed with all other plates in the batch These differences may manifest themselves
11. Pre genotyping Performance with DQC Box Plots Monitoring DQC plate box plots Figure 5 3 is an effective method for early flagging of problematic plates and detecting trends in plate performance because DQC is a single sample metric that is computed early and quickly for every sample on every plate Step 2 Generate Sample DOC Values on page 15 Laurie CC Doheny KF Mirel DB Pugh EW Bierut LJ Bhangale T Boehm F Caporaso NE Cornelis MC Edenberg HJ Gabriel SB Harris EL Hu FB Jacobs KB Kraft P Landi MT Lumley T Manolio TA McHugh C Painter I Paschall J Rice JP Rice KM Zheng X Weir BS GENEVA Investigators Quality control and quality assurance in genotypic data for genome wide associ ation studies Genet Epidemiol 2010 Sep 34 6 591 602 Chapter 5 Additional Sample and PlateQC 33 A suggested approach is for each plate of samples create a box plot of the DQC values arrange them in chronological order and identify the median DQC value for each plate Next identify the median of the DQC medians and the standard deviation for each array plate Finally identify any plates whose 25th percentile upper bound of the box is lower than 2 standard deviations below the median of medians Such outlier plates should be flagged for further consideration especially if the boxplot is visually obviously much lower than the rest of the plates We note that being an outlier by this 2 standard deviation definition does not n
12. QC genotyping with SNP specific models file if the batch size is less than 96 samples Figure 7 5 Performing genotyping from the Workspace pop out menu _B DataSets gt i A Sample Attributes gt E Show Intensity QC Table 41 Genotype Results HH Show Signature Genotypes Reports gt Copy Number LOH Results gt Copy Number Variation Results SNP Lists Copy Number LOH Analysis Perform Copy Number Variation Ang sis Properties Select a user intensity data group All Out of Bounds Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 46 a Right click the In Bounds data group displayed in the data tree and select the Perform Genotyping option Figure 7 6 n Choose the 96orMore_Step1_AxiomGTT1 as the analysis configuration in the drop down menu if the batch size is equal to or more than 96 samples Refer to Figure 7 4 for more details n Choose the LessThan96_Step1_AxiomGT1 as the analysis configuration from the drop down menu to perform QC genotyping with SNP specific models file if the batch size is less than 96 samples Figure 7 6 Performing genotyping by right clicking the In Bounds data group in the data tree and selecting the Perform Genotyping command from the drop down menu B Workspace Axiom Gw LAT c Data Sets c i Axom Gw LAT Axom KP UCSF LAT E Sample Attributes S38 Inte
13. as putative differences in the MAF of SNPs over these samples relative to the rest of the study set If such a plate effect is also combined with a poor study design where cases or controls are genotyped separately on different plates this may greatly increase the false positive rate in the GWA study Even in a well designed study where cases and controls are randomized across plates inclusion of such outlier plates will decrease the power and or increase false positive rates Chapter 3 Best Practices Genotyping Analysis Workflow 17 The metrics and guidelines for plate performance are as follows Metrics Samples passing DOC and 97 QC call rate X 100 Total samples on the plate m Plate pass rate m Average QC call rate of passing samples on the plate MEAN QC call rates of samples passing DQC and 97 QC call rate thresholds Guideline for High quality Plates m Plate pass rate gt 95 for samples derived from tissue blood or cell line and gt 93 if sample source is saliva m Average QC call rate of passing samples gt 99 Guideline for Passing Plates m Average QC call rate of passing samples gt 98 5 The minimum guideline for passing plates is an average QC call rate of passing samples that is greater than or equal to 98 5 gray and green zones in Figure 3 2 Ideally all plates in the batch will pass the guidelines for high quality plates green zone Figure 3 2 Passing plates in the gray zone shou
14. axiom_array gt _LessThan96_Step2 r lt gt apt N A Required for small sample size lt axiom_array gt r lt gt AxiomGT1 MODELS Required for small sample size Required for small sample size Chapter 1 Introduction to Axiom Data Analysis 8 Table 1 1 Files Used For Analysis of Axiom Genotyping Arrays axiom array will be replaced with the actual name of the array lt R gt will be replaced with the version of the analysis files For example when lt axiom_array gt Axiom BioBank1 and lt R gt r2then axiom array 96orMore Step2 r it apt probeset genotype AxiomGT1 xml2 Axiom BioBank1 96orMore Step2 r2 apt probeset genotype AxiomGT1 xml Continued Analysis Library Files GTC APT lt axiom_array gt r lt gt apt geno QC AxiomQC1 xml N A Optional lt axiom_array gt r lt gt step2 ps Optional Optional lt axiom_array gt apt probeset N A a Optional for human genomes genotype AxiomSS1 xml a N A for non human genomes lt axiom_array gt r lt gt signatureSNPs ps a Required for human genomes a Optional for human genomes a N A for non human genomes a N A for non human genomes lt axiom_array gt r lt gt psi Required N A lt axiom_array gt array_set Required N A lt axiom_array gt AxiomGT1 gc_analysis_parameters Required N A lt axiom_array gt gc_analysis_configuration Required N A lt axiom_array gt analysis Required N A lt axiom_array gt qc_thresholds Required N A
15. be excluded from step 7 genotyping enclosed in circles Plate metrics for a batch of 39 Plates o e 2 2 High Quality Passing Plates Plates Mean QC e ui Call Rate 99 06 of Passing Samples amp on Plate Plate Pass Rate This section describes minimum required Plate QC step See Additional Plate OC on page 32 for additional Axiom specific methods and general methods used in the field to detect outlier plates and batches Step 7 Genotype Passing Samples and Plates Over Step2 AxiomGT1 SNPs For this step all samples in the batch that passed sample QC and Plate QC Steps 3 5 and 6 are co clustered and genotype calls are produced by the AxiomGT1 algorithm This Best Practices Step 7 is referred to as Step2 AxiomGTI genotyping in the instructions provided for genotyping with GTC Chapter 7 and APT Chapter 8 Depending on the array Step2 AxiomGTI genotyping produces calls for all probe sets on the array or only a subset Probe sets excluded by Step2 AxiomGT genotyping are usually those with repeatable performance problems and or genetic complications As discussed in What is a SNP Cluster Plot for AxiomGTI Genotypes on page 10 the AxiomGT1 algorithm can be executed with generic priors or SNP specific priors The best practice recommendation is to use SNP specific priors for small batches lt 96 samples Use of generic priors is generally recommended for large batches gt 96 samples whe
16. comprised of at least five different study samples and if any of the specific samples or plates are poorly performing outliers they should be removed from use in the reproducibility test If this quantity and variety of replicates are not available reproducibility can still be used as a coarse filter for SNPs with obvious low values Instructions for Executing Best Practices Steps with GTC and APT Software This section provides an overview of the QC and genotyping workflows to be used in Affymetrix Genotyping Console GTC version 4 2 or higher and Affymetrix Power Tools APT version 1 16 1 or higher Execute Steps 1 7 with GTC Software GTC Setup Analysis library files and annotation files can be directly downloaded from within GTC 4 2 or they can be manually downloaded from www affymetrix com and unzipped into the current GTC library folder For more detailed instructions on how to install GTC and obtain analysis library and annotation files please consult the Genotyping Console User Guide P N 702982 available at the support section of the Affymetrix website www affymetrix com Step 1 Group Samples into Batches in GTC Samples can be grouped into batches either within or outside GTC In order to group samples into different batches within GTC import all CEL files from the study into a dataset and a then add the selected CEL files to individual custom intensity data groups batches in the Intensity QC table
17. correspondence between probe set and SNP site some SNP sites are interrogated by more than one probe set The example script also includes options for additional output files The posteriors and summary file must be created for use in Step 8 The AxiomGTIl snp posteriors txt file is enabled by write models option and includes the location and variance of the genotype clusters per probe set a The AxiomGTI1 summary txt file is enabled by summaries option and includes the summarized intensity for the A and B allele of each probe set and sample CHP files one for each sample are enabled by the cc chp output option CHP files can be input into GTC and thus enable analyses such as viewing of cluster plots and generation of files in PLINK format Note Choose lt axiom_array gt _LessThan96_Step2 r lt gt apt probe set genotype AxiomGT 1 xml to perform genotyping with SNP specific models if batch size is less than 96 samples Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 65 Execute Best Practice Step 8 with SNPolisher Functions SNPolisher Setup The R package files to install SNPolisher are available on the Affymetrix website www affymetrix com Select Register at the top of the website to register your email address with Affymetrix Once you have registered SNPolisher can be downloaded from the DevNet Tools page From the Partners and Programs menu select Developers Network L
18. e 2 i 0 1 2 2 1 0 1 2 A B A B A A SNP plotted with reference genotypes B The same SNP plotted without reference genotypes priorFile takes the name of a file containing the prior information for the genotypes default is NULL This file must be in the same format as the posterior information file The accompanying logical operator plot prior indicates if the prior genotype cluster centers are plotted default is FALSE If plot prior is FALSE Ps_Visualization ignores priorFile If plot prior is TRUE and priorFile is NULL then Ps_Visualization plots a generic prior Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 69 match cel file name is a logical operator indicating if sample file names in the calls file are checked against those in the confidence summary files Input files may not have been checked against each other so the default is TRUE max num SNP draw is the maximum number of SNPs that should be plotted If the list of probe set IDs used is one of the categorical lists output by Ps Classification it is important to set max num SNP draw we suggest lt 500 or all SNPs in the list will be plotted nc us is the number of genotype clusters The default value is 3 and needs to be set of 5 for auto tetraploids geno col is the list of colors for plotting the genotypes The input for geno col must be given as a vector of colors To make a vector use the command c for concatenate
19. first line should read sample color separated by a tab To keep the highlighted samples colored and change the color of all other samples to a single color set geno col to be only one color geno col c lightgray see Figure 8 4 B When using multiple colors for highlighted samples be sure to check that the colors are clearly visible against the colors used for other samples Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 70 Figure 8 4 One SNP with Samples Highlighted in Two Colors A B ps101 ps101 AvM called genotypes AvM called genotypes e ei c e e E lt lt N 2 1 0 1 2 2 1 0 1 2 A B A B A One SNP plotted with default color options B The same SNP with all non highlighted and samples highlighted in green and turquoise samples plotted in light gray If the user in the previous examples wishes to produce cluster plots after running OTV Caller the command in R should be gt Ps Visualization pidFile PolyHighResolution ps output pdfFile Cluster PolyHighResolution pdf summaryFile C data AxiomGT1 summary txt callFile C data AxiomGT1 calls txt confidenceFile C data AxiomGT1 confidences txt posteriorFile C data AxiomGT snp posteriors txt temp dir Temp keep temp dir FALSE refFile NULL plot prior T priorFile NULL atch cel file name TRUE max num SNP draw 6 99 66 99 66 99 66 geno col c red go
20. lower DNA quality and or contaminated samples Genet Epidemiol 32 676 Cardon LR Palmer LJ Population stratification and spurious allelic association Lancet 2003 Feb 15 361 9357 598 604 3 Clayton DG Walker NM Smyth DJ Pask R Cooper JD Maier LM Smink LJ Lam AC Ovington NR Stevens HE Nutland S Howson JM Faham M Moorhead M Jones HB Falkowski M Hardenbol P Willis TD Todd JA Population structure differ ential bias and genomic control in a large scale case control association study Nat Genet 2005 Nov 37 11 1243 6 Zondervan KT Cardon LR Designing candidate gene and genome wide case control association studies Nat Protoc 2007 2 10 2492 501 Chapter 3 Best Practices Genotyping Analysis Workflow 14 Execute the Required Steps of the Workflow This section describes the eight steps that are required for the Best Practices Analysis Workflow and recommended for all Axiom Genotyping Arrays Figure 3 1 Step 1 Group samples into batches For each batch perform the following Step 2 Generate sample DQC values Step 3 QC samples based on DQC values Step 4 Generate sample QC call rates Step1 AxiomGT1 Step 5 QC samples based on QC call rate Step 6 QC the plates offline Excel or script Step 7 Genotype passing samples amp plates over Step2 AxiomGT1 SNPs Genotype CALLS table 4 f Legend N ee Step completed in Affymetrix Power Tools or Genotyping Console Step completed
21. lt axiom_array gt gt_thresholds Required N A lt axiom_array gt geno_intensity_report Required N A Chapter 1 Introduction to Axiom Data Analysis 9 Introduction The success of a genome wide association study GWAS in finding or confirming the association between an allele and disease and traits in human plant and animal genomes is greatly influenced by proper study design and the data analysis workflow including the use of quality control QC checks for genotyping data Although the number of replicated allele complex disease associations discovered through human GWAS has been steadily increasing most of the variants detected to date have small effects and very large sample sizes have been required to identify and validate these findings As a result even small sources of systematic or random error can cause false positive results or obscure real effects This reinforces the need for careful attention to study design and data quality In addition most genotyping methods assume three genotype clusters AA AB BB for two alleles This assumption does not always hold especially in plant and animal studies due to the existence of subpopulation genome structural variation and or auto polyploid genomes This guide presents the Best Practices Genotyping Analysis Workflow to address these challenges along with instructions for using Axiom software for all human plant and animal Axiom Genotyping Arrays The Axiom Genot
22. may also be appropriate to examine as part of the SNP filtering process Hardy Weinberg p values pHW are computed by GTC No Axiom software is provided for reproducibility calculation and Mendelian trio error For these additional metrics absolute QC and pass fail thresholds can only be set in the context of the study design The general guideline is to examine the distribution of each metric and then examine cluster plots for SNPs with outlier values and over a collection of randomly selected SNPs Thresholds may be set based on consideration of three properties a the absolute value of the metric the deviation from the mean median values and a the expectation based on an examination of cluster plots that SNPs below a threshold are likely to be misclustered Hardy Weinberg p value The Hardy Weinberg p value pHW is a measure of the significance of the difference between the observed ratio of heterozygote calls in a population and the ratio expected if the population is in Hardy Weinberg equilibrium HWE The test should be performed on unrelated individuals with relatively homogenous ancestry Although genotyping artifacts may produce low pHW values using this as a SNP QC metric can be tricky because a low p value may be caused by true genotypic frequency deviation Examination of cluster plots indicates that most of the extreme deviations p value lt 107 are due to poorly performing SNPs In GTC the SNP summary statistics
23. temporary directory if the intermediate output files are needed for closer inspection Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 68 Ps_Visualization has eight optional arguments sampFile takes the name of a file containing a list of samples to be highlighted in a plot This file has no header line and the CEL or sample names must match the names in the summary and calls files refFile takes the name of a file containing reference genotypes for plotting default is NULL The source of reference genotypes may be a SNP discovery project such as HapMap or the 1000 genomes project or it may be genotypes produced in an NGS project for the user s samples The user must create the refFile with the same format and genotype codes as the AxiomGT1 calls txt file Samples without reference genotypes have their reference plotted as a gray square for No call If there are very few reference genotypes the reference plot will contain mostly gray squares In this case the user may wish to consider plotting only those samples with known reference genotypes Figure 8 2 Figure 8 2 One SNP Plotted With and Without Reference Genotype ps101 ps101 AvM called genotypes AvM called genotypes o o e e e t e oe ao o ES lt x ae e N 2 1 0 1 2 a 1 0 1 2 A B A B ps101 ps101 AvM reference genotypes AvM reference genotypes o o S e m Q e o 9 a 9 t lt lt x x e
24. unique signatures in the genotyping samples match the count expected in the study set and that the signatures of expected replicates are the same and are found in the expected plate positions In addition a check that the called genders match the expected genders for each sample is recommended Unusual or Incorrect Gender Calls Samples with either unusual or incorrect gender calls as determined by comparing the reported gender for each sample with the actual gender and or by comparing the genders of repeated samples should be carefully examined before they are included in analyses Methods for checking gender and detecting sex chromosome aneuploidy are presented in Laurie et al Detecting Mixed Contaminated DNA samples This section discuss patterns produced by mixing of genomes from multiple individuals The more of these patterns that occur for a sample the more likely it is that contamination is the causal factor However since contamination is not the only cause of these patterns ultimately the investigator s judgment is required to determine whether these samples should be included in further analyses Samples Have Relatively High DQC and Low QC Call Rate QCCR Values In general higher DQC values correlate with higher sample call rates see Figure 5 1 A one exception is when samples are contaminated DQC values are produced by non polymorphic probes and so are not sensitive to the mixing of DNA from different individuals Howe
25. variants in genotyping arrays improves genotype retention and reduces ascertainment bias BMC Genomics 2012 Jan 19 13 34 Chapter 4 Additional Genotyping Methods 29 Genotyping Auto tetraploids Auto polyploids occurring in some plant and fish species are polyploids with two sub genomes generally derived from different species SNP sites have a maximum of 6 possible genotypes AA AA AA AB AA BB BB AB BB BB AB AB and 5 intensity clusters AA BB cannot be distinguished from AB AB Because AxiomGT1 genotypes a maximum of three genotype clusters the workflow for assigning genotype calls for auto tetraploid genomes is different from the workflow for allo polyploid and diploid genomes The R package fitTetra http cran r project org web packages fitTetra index html produces genotypes for auto tetraploids and is recommended for Axiom arrays designed to interrogate such genomes fitTetra was developed by Dr RE Voorips at Wageningen University s Plant Breeding section The paper describing the fitTetra algorithm is available Voorips et al 2011 SNPolisher functions provide a workflow to 1 Generate the needed Axiom data 2 reformat Axiom data for fitTetra input 3 use fitTetra R package for assigning genotype calls and then 4 reformat fitTetra output for use of SNPolisher functions on the produced calls See Section 3 8 of the SNPolisher User Guide for detailed descriptions of the fitTetra input and output functions as well
26. with Excel or R script a Step completed with SNPolisher Functions or APT version 1 16 1 or higher A The actual commands used to execute the steps differ between the software tools APT GTC and SNPolisher Instructions for using GTC to execute the Best Practices Workflow Steps 1 7 followed by APT for Step 8 are provided in Chapter 7 Instructions for using APT to execute Best Practices Workflow Steps 1 7 followed by SNPolisher functions for Step 8 are provided in Chapter 8 Step 1 Group Sample Plates into Batches In general group plates in as large a batch size as is computationally feasible or up to 50 plates in the order in which the plates were processed e g if using batches of 8 plates it is usually preferable to group together the first 8 the second 8 etc The minimum batch size when using generic priors is 96 samples comprising at least 90 unique individuals Chapter 3 Best Practices Genotyping Analysis Workflow 15 SNP specific priors should be used when the total batch size is between 20 and 96 unique individuals The specific genotyping option for large 7 96 samples or small 96 samples batch sizes must be chosen in both GTC and APT workflows Each batch should contain either 15 or more distinct female samples or zero female samples In other words if any female samples are going to be genotyped at least 15 distinct female samples must be included in the batch The exceptions to these batching r
27. 04 8 309 496 8 501 122 and other U S and foreign patents The array imaging system utilized in the Axiom Genotyping Solution may be covered by one or more of the following patents U S Patent Nos 5 578 832 5 631 734 5 834 758 5 981 956 6 025 601 6 141 096 6 171 793 6 207 960 6 225 625 6 252 236 6 490 533 6 511 277 6 604 902 6 650 411 6 643 015 6 741 344 6 789 040 6 813 567 7 062 092 7 108 472 7 130 458 7 222 025 7 689 022 7 983 467 7 992 098 8 208 710 8 233 735 8 391 582 and other U S and foreign patents The software products utilized in the Axiom Genotyping Solution may be covered by one or more of the following patents U S Patent Nos 5 733 729 5 795 716 5 974 164 6 066 454 6 090 555 6 185 561 6 188 783 6 223 127 6 228 593 6 229 911 6 242 180 6 308 170 6 361 937 6 420 108 6 484 183 6 505 125 6 510 391 6 532 462 6 546 340 6 567 540 6 584 410 6 611 767 6 687 692 6 607 887 6 733 964 6 826 296 6 882 742 6 957 149 6 965 704 6 996 475 7 068 830 7 130 458 7 215 804 7 424 368 7 634 363 7 822 555 7 991 564 7 992 098 8 190 373 8 498 825 and other U S and foreign patents Copyright 2011 2014 Affymetrix Inc All rights reserved Contents Chapter 1 Chapter 2 Chapter 3 Chapter 4 Introduction to Axiom Data Analysis 0000 cece eee eee 6 Aboubthis GUINE ccs gece ee ce ce E tn tee ten REI en cig UE E E AE A 6 dico c RTT 6 Prerequisites sss nian eam pee pu RR RR I
28. 0S H 10AxomGT1 cho female Export Genotype Resuks for PLINK Copy Number LOH Results Copy Number Variation Results i1 Reports LJ SNP Lists Input Value Enter a name for the intensity data group 20110923 165605 CEL outer excluded after QCgenotyping 3 Workspace Axiom Gw LAT lf CHF e Data Sets D S E Axom Gw LAT Default v MU Axiom KP UCSF LAT E Sample Attributes Ge Intensity Data gt 1 T All 5 lf In Bounds c li Out of Bounds 3 s Create Custom Intensity Data Group Excluding Selected S ample eB epe Results HE Show Intensity QC Table 20110923 161145 CHP st f Copy Number LOH Results Show Signature Genotypes Copy Number Variation Results Perform QC Reports E 3 SNP Lists i Perform Genotyping Copy Number LOH Analysis gt Perform Copy Number Variation Analysis Rename Intensity Data Group Remove Intensity Data Group LI Jb Select Gender File LI None Hirks Fie v Select Output Root Path C Ada LAT CEL lal Select Base Batch Name 20111123_150045_CHP step2 Output File Suffix leave blank for no sulfa 7 Use Defaut Base Batch Name as Posterior Models Fie Name Select Posterior Models File Name postenior models 20111123_150045_CHP step2 Co coxa J Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 49 Execute Step 8 with APT Version 1 16 1 or Higher In this section we provide instructions fo
29. 24 Allo polyploid SNP Cluster Plot Pattern 2 0 2 eee 26 SNP Cluster Plot Patterns for Inbred Populations 00 0000 e eee 27 Additional Genotyping Methods 0 000 cece eee eee eens 28 Manually Change Genotypes 0 0 0c eee ee 28 Adjust Genotype Calls for OTV SNPs 0 eee ee 28 Genotyping Auto tetraploids 2 0 eee eae 29 Increase the Stringency for Making a Genotype Call l a 0 0 0 0 20 00 0000 else 29 Contents 4 Chapter 5 Additional Sample and Plate QC 0 060 ee 30 Additional Sample QC 22 4552 2duee42464a n4 540060 2451449 bom dde been idee 30 Detecting Sample Mix ups nana aaa anaa 30 Unusual or Incorrect Gender Calls lille 30 Detecting Mixed Contaminated DNA samples 0 0 0000 0 eee eee 30 Samples Have Relatively High DQC and Low QC Call Rate QCCR Values 30 Samples Have a High Percentage of Unknown Gender Calls lusu 31 Samples Tend to Fall Between the Genotype Clusters Formed by the Uncontaminated Samples 00 000 00 e 31 Unusual Patterns of Relatedness llle 32 Increased Computed Heterozygosity 0 0 0 eee 32 Additional Plate QC oo eee eee 32 Evaluate Pre genotyping Performance with DQC Box Plots 000 00s 32 Monitor Plate Controls 2 0 s 33 Check for Platewise MAF Differences llle 34 Chapter 6 SNP a a ca Rariaiu aee Sepe iNcirecy iut ER f 35 SNP Metrics Used in the Ps Classification S
30. Correcting a Cluster Split AX 13259111 cc chp Other_rand200 Genotype AA Was ABB oTv 4 NoCall Ellipses OBB QAB QAA QB QA Prior N Posterior 11 44 10 06 Log Ratio AX 13259111 cc chp Other_rand200 12 15 or UN Genotype A QAA a 7 Has an Ba ABB tg arte 4FoTV z Ne a X3 N PY P NoCall B Ellipses OBB QAB QAA QB OA Prior N Posterior 11 45 Fi aaf x 10 05 imd p El dod T 9 2 2 2 22 2 7 s ren a EI 9 35 4 24 254 0 84 0 87 257 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 59 Saving a Cluster Plot There are several different methods for saving a cluster plot a To copy the cluster plot to the clipboard right click the plot and select Copy image to clipboard Figure 7 18 A To save the cluster plot as a PNG file right click the plot and select Save image to file Figure 7 18 B Save image to file Create Custom Intensity Data Group From Checked Samples Create Custom Intensity Data Group Excluding Checked Samples Scale To save the cluster plots for all SNPs in a SNP List to a single PDF file click the Save All Cluster Graphs to PDF shortcut on the SNP Cluster Graph tool bar Figure 7 19 te Cluster Graph 20090508 134344 The first page of the PDF displays the common lege
31. GACTCTT C A TTTTTIATTCGTIGAACCTAGAGGGGGTCA MM M Reverse strand probeset Marker A marker refers to the genetic variation at a specific genomic location in the DNA of a sample that is being assayed by the Axiom Genotyping Solution Both SNPs and indels can be genotyped The Affymetrix unique identifier for a marker is referred to as an affy snp id An affy snp id is comprised of the prefix Affx followed by an integer for example Affx 19965213 A set of one or more probe sequences whose intensities are combined to interrogate a marker site is referred to as a probe set Most Axiom markers are interrogated with one or two probe sets one derived from the forward strand sequence and or one derived from the reverse strand sequence The Axiom identifier for a probe set is referred to as a probeset id A probeset id is comprised of the prefix AX followed by an integer for example AX 33782819 For simplicity in this document the term SNP is used to refer to both SNPs and indels In addition the term SNP is often used to as shorthand for the probe set used to interrogate the SNP or indel What is a SNP Cluster Plot for AxiomGT1 Genotypes A SNP cluster plot corresponds to one probe set designed to interrogate a given SNP and each point corresponds to one sample whose A and B allele array intensities have been transformed into the X vs Y coordinate space used by the AxiomGT genotyping cluster algorithm Functions for creati
32. LAT_CELs 20110923_165853_CHP step2 K NDNA01006_B04_K NDNA01006_B_ 4 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_B05_K NDNA01006_B_5 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_B06_K NDNA01006_B_6 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNAO1006_B07_K NDNA01006_B_7 AxiomGT 1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_B08_K NDNA01006_B_8 AxiomGT1 chp Files 576 B Use the file location data to identify the folder where the files are stored 20110923_165853_CHP step2 File Edit View Favorites Tools Help Q sxx gt 27 JO search Py Folders E Address C data LAT_CELs420110923_165853_CHP step2 Right click the dataset and select Show File Locations AxiomGT1 Genotyping Console log AxiomGT1 report txt 55 AxiomGT1 snp summary db K NDNA01006 A02 K NDNA01006 A 2 AxiomGT1 chp Size Type 319 896 KB Text Document 3KB Text Document S9KB Text Document 56 804 KB Data Base File 36 065 KB CHP File C Navigate to the folder and confirm that the Step2 AxiomGT1 Step 7 files are all there Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 51 Step 8A Run Ps_Metrics Example Ps_Metrics Script In this example the working directory is input And the output is written to input metrics txt The Ps Metric program is available in APT versions 1 16 1 and higher Ps Me
33. LINK compatible formats and directly analyze the output files using PLINK software Figure 7 21 Exporting genotype results from GTC for third party analysis Workspace Axiom_GW_LAT cH e Data Sets gi Axiom Gw LAT Axiom KP UCSF LAT Ej Sample Attributes gt 4 38 Intensity Data Qs Al 2 T In Bounds 3 liy Out of Bounds 4 py Create Custom Intensity Data Group Excluding Selected Samples S E Genotype Results T 20110923_161145_CHP step1 20110923_165853_CHP step2 a V Copy Number LOH Results Show CHP Summary Table Copy Number Variation Results Reports Show SNP Summary Table L SNP Lists Create SNP List Show SNP Cluster Graphs Create Custom Intensity Data Group Export SNP Summary Table Export Merged Genotype Results Run CHP vs CHP Concordance Check Run CHP vs TXT Concordance Check Rename Genotype Results Group X Remove Genotype Results Group The Export Genotype Results and Export Genotype Results for PLINK commands are found in the pop out menu that appears after right clicking a genotyping results batch in the data tree 8 Instructions for Executing Best Practices Steps with Command Line Software Execute Best Practice Steps 1 7 with APT Software In this section we provide instructions for executing steps 1 8 of the best practice analysis workflow see Figure 3 1 using APT combined with some simple scripts
34. No Call assignment N the number of samples over which a genotype call is attempted for the SNP SNP Call Rate CR is the ratio of the number of samples assigned a genotype call of either AA BB or AB for the SNP i e the number of samples that do not have No Call to the number of samples over which a genotype call is attempted for the SNP SNP call rate is a measure of both data completeness and genotype cluster quality at low values Very low SNP call rates are due to a failure to resolve genotype clusters Figure 6 1 A Poor cluster resolution may produce inaccurate genotypes in the samples that are called or a non random distribution of samples with no calls and may lead to false positive associations in a GWA study Figure 6 1 SNPs with Different SNP Call Rates CR A SNP with low 93 096 CR B SNP with high 99 496 CR 11 0 10 0 10 5 10 5 11 0 11 0 9 5 Although SNP Call Rate is correlated with genotype quality the performance of marginal SNPs falls along a continuum and there is no perfect threshold for filtering out problematic SNPs from a pool of SNPs providing optimal power for a study We recommend setting the filtering thresholds for CR based on the species under study and visually examining the cluster plots for SNPs with CR just above or below the threshold This examination may result in the inclusion of some SNPs with CR just below the threshold as well as the removal of some SNPs with CR ju
35. Other ps and OTV ps will be used with the GTC SNP Cluster Graph function The default FALSE In order to visualize probe sets listed in the category files with the GTC SNP Cluster Graph function these files must be imported into GTC as a SNP list The GTC SNP list file must be a text file with a txt or tsv extension and contain a column labeled Probe Set ID If the GTC flag in SNPolisher Ps_Classification was set to FALSE the column header must be updated by hand Open the category file in a text editor such as Notepad or Word and change probeset_id to Probe Set ID then save with a txt extension If the GTC flag in Ps_Classification was set to TRUE the category files still must be renamed with a txt or tsv extension Ps Classification accepts a list of probe sets and will categorize the SNPs in this file only The first row of this file should always be probeset 1d See the SNPolisher User Guide for longer examples of running the SNPolisher functions and for more details of the arguments for Ps Classification Note that if SNPolisher v 1 4 or less is used the example command will be gt Ps Classification metrics file metrics txt output dir Output SpeciesType Human ps2snp file XANALYSIS FILES DIR gt Axiom_BioBank1 r2 ps2snp_map ps Visualize SNP Cluster Plots with SNPolisher Ps Visualization Function Once the Ps Metrics Ps Classification and OTV Caller functions have been run the Ps Vis
36. Q vs QC Call Rate QCCR Plots A Representative data set of 10 plates with no obvious contamination problems B One plate including 4 samples enclosed in box where DNAs were deliberately mixed C Five plates one plate green contains samples that were accidentally contaminated during pipetting 101 100 5 100 i 100 4 o AMG 99 4 995 gt 9 8 99 4 c i 499 9 e s 98s 8 97 eG s 26 j 94 96 0 88 0 9 0 92 0 94 0 96 0 98 1 955 7 T 0 93 0 94 0 95 0 96 0 97 0 98 0 99 Dac DQC o e 8 7 as x Oo a o o 8 e u o v 0 80 085 0 90 0 95 1 00 DQC Samples Have a High Percentage of Unknown Gender Calls If male and female DNA are mixed in high enough proportions the Axiom gender calling algorithm will set the call to unknown Note that individuals with unusual genders for example XXY will also tend to have gender unknown calls Samples Tend to Fall Between the Genotype Clusters Formed by the Uncontaminated Samples The cluster plots in Figure 5 2 include deliberately mixed samples red and these points fall between the cluster locations for pure BB AB and AA genotypes See the SNPolisher User Guide and usage of Ps Visualization for instructions to color specific samples in a cluster plot Chapter 5 Additional Sample and PlateQC 32 Figure 5 2 Cluster plots for two SNPs Deliberately mixed samples are colored red Uncontaminated samples
37. Software 50 Figure 7 9 How to locate the Step2_AxiomGT1 Step 7 output files Data Sets A UCSF LAT Axom E Create Data Set tz c Remove Data Set Data Set UCSF LAT DataSet Files Genotype Results Files UCSF LAT 20110923 165853 CHP step2 NDNA01006 A02 K NDNAO1006 A 2 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A03_K NDNA01006_A_3 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A04_K NDNA01006_A_4 AxiomGT 1 chp CdataALAT CELsN20110923 165853 CHP step2W NDNA01006 A05 K NDNAO01006 A 5 AxiomGT 1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A06_K NDNA01006_A_6 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A07_K NDNA01006_A_7 AxiomGT 1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A08_K NDNA01006_A_8 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A09_K NDNA01005_A_3 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNAO1006_A10_K NDNA01006_A_10 AxomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_A11_K NDNA01006_A_11 AxsomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNAO1006_A12_K NDNA01006_A_12 AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_B01_K NDNA01006_B_1 AxiomGT 1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNA01006_B02_K NDNA01006_B_2AxiomGT1 chp C data LAT_CELs 20110923_165853_CHP step2 K NDNAO1006_B03_K NDNA01006_B_3 AxiomGT1 chp C data
38. User Guide for more details on the priority order argument for Ps Classification The file lt axiom_array gt r lt gt ps2snp_map ps in the Analysis Library File package Table 1 1 on page 7 contains the list of matched probe sets and SNPs Chapter 3 Best Practices Genotyping Analysis Workflow 20 Poly High Resolution Mono High Resolution No Minor Homozygote Ei z E 2 E 3 2 10 1 2 3 ow 4 Yy 2 3 3 2 1 0 1 2 3 Good duster resolution High absolute Contrast value Good cluster resolution At least 2 examples of minor allele All genotyped samples are monomorphic No minor homozygous examples Off Target Variant Call Rate Below Threshold Other e s T 3 ve z 3 s 3 gt L a 3 2 1 0 1 2 3 2 1 0 1 2 woo 0 2 4 Has on Off Target Variant cluster Call Rate is below threshold but all One or more cluster properties are arrow other cluster properties are good below threshold values Hemizygous 114 105 110 3 mite z 3 2 1 0 1 2 3 E 2 0 2 Y SNP Mito SN females are set to No Call arrow Chapter 3 Best Practices Genotyping Analysis Workflow 21 Table 3 1 Default QC Thresholds for CR FLD HetSO HomRO metrics defined in Chapter 6 and nMinorAllele number of minor alleles in the batch Metric values must be greater than or equal to the threshold in order to be considered passing HetSO OTV is the HetSO threshold for OTV detection see Adjust Genotype C
39. _Classification software described above follows these conventions and can be imported directly into GTC The SNPolisher Ps_Classification function also outputs the 7 category files named PolyHighResolution ps NoMinorHom ps Hemizygous ps MonoHighResolution ps CallRateBelowThreshold ps Other ps and OTV ps These SNPolisher files require manual changes before they can by imported into GTC The default column header for Ps_Classification category files is probeset_id and the default file extension is ps If the GTC flag in SNPolisher Ps Classification was not set to TRUE the column header must be updated by hand Open the category file in a text editor such as Notepad or Word and change probeset_id to Probe Set ID then save with a txt extension If the GTC flag in Ps Classification was set to TRUE the category files still must be renamed with a txt or tsv extension To Import a SNP List 1 Right click SNP Lists in the data tree and select Import SNP List see Figure 7 10 Figure 7 10 Select SNP List ij Show Information 2 The Open dialog box appears Navigate to the location of the SNP List and select a list Click Open see Figure 7 11 Figure 7 11 The Open dialog box Open Look in 9 Checked SNPs E Best 5 Checked SNPs TSV My Recent Documents Desktop My Documents My Computer File name My Network Files of type TSV Files TSV Chapter 7 Instructions
40. alls for OTV SNPs on page 28 HomRO1 HomRO2 and HomRO3 are the HomRO thresholds for SNPs with 1 2 or 3 genotypes respectively nMinorAllele is the threshold for the minimum number of minor alleles in order for a SNP to be classified as PolyHighResolution For more information see the SNPolisher User Guide Metric Human Diploid Polyploid CR 95 97 97 FLD 3 6 3 6 3 6 HetSO 0 1 0 1 0 1 HetSO OTV 0 3 0 3 0 3 HomRO1 0 6 0 6 N A HomRO2 0 3 0 3 N A HomRO3 0 9 0 9 N A nMinorAllele 2 2 2 The Ps Classification function outputs the Ps performance txt file which contains the probeset id s affysnp id s QC metrics hemizygous status and an indicator if this probe set is the best for the SNP BestProbe set and which classification Figure 3 3 the probe set belongs to ConversionType for each probe set If all SNPs have one probe set then every probe set is the best probe set by default Column Names and Examples are shown below Table 3 2 Table 3 2 Ps performance Column Names and Examples Column Name Example probeset id AX 11481545 affy snp id Affx 27771153 CR 99 232012934519 FLD 8 19369622294609 HomFLD 17 9734932365231 HetSO 0 450052256708187 HomRO 2 57169 nMinorAllele 5123 Nclus 3 n AA 3112 n AB 3383 n BB 870 n NC 57 hemizygous 0 HomHet 0 ConversionType PolyHighResolution BestProbeset 1 Chapter 3 Best Practices Gen
41. amples of multi cluster SNPs A Diploid plant SNP with 7 8 intensity clusters B Diploid plant SNP with 4 5 intensity clusters one is an OTV 2 e o E o o o So o eo wo o OTV N 72 1 0 1 2 c 4 0 1 2 3 A B Chapter 3 Best Practices Genotyping Analysis Workflow 26 Allo polyploid SNP Cluster Plot Pattern Allo polyploid genomes contain more than two paired sets of chromosomes where each set is referred to as a sub genome and the sub genomes are derived from different species The alleles of allo polyploid SNP sites usually segregate in just one sub genome while remaining fixed in the homeologous sites in the other sub genomes Allo polyploid genomes occur in some plant and fish species and produce expected differences in SNP cluster patterns Figure 3 7 relative to diploid genomes Figure 3 5 left The intensity contributions of fixed sub genomes do not create additional clusters but they shift and compress the clusters formed by the sub genome with the segregating alleles to the right when A is the fixed allele or left when B is the fixed allele The heterozygous genotype cluster is located between the homozygous genotype cluster along the Y Size axis The AxiomGT1 genotyping algorithm dynamically adapts to the shifted cluster locations and allo polyploid SNPs with the expected pattern are classified as PolyHighResolution when the Polyploid option is selected in the Ps_Classification step see SNPolisher User Guide fo
42. an three intensity clusters may be produced Figure 3 6 This multi cluster pattern may be due to genuine genetic differences in the clustered samples especially when genotyping plant and animal genomes or the pattern may be an artifact due to extreme batch effects Batch effects variables include sample collection source plate ID instrument operator sample type processing date and more Possible genetic differences may be due to inclusion of subpopulations with copy number variations at the given SNP site or inclusion of subpopulations whose genomes have diverged from the reference population whose genome sequence was used to design the probes for the array Genomes of divergent subpopulations may have interfering SNPs and indels relative to the array probe sequences that decrease the genotype intensities OTV SNP sites Didion ef al 2012 are extreme cases where genomes have diverged to the point where only background intensities are produced and a fourth intensity cluster is formed at the het cluster position An example is shown in Figure 3 6 B The AxiomGT1 genotyping algorithm assumes a maximum of three genotype clusters for just two alleles and thus will merge additional intensity clusters into three genotype states resulting in unpredictable mis calling of the true complex genetic states Didion JP Yang H Sheppard K Fu CP McMillan L de Villena FP Churchill GA Discovery of novel variants in genotyping arrays improves geno
43. ant FLD HomFLD Heterozygous Strength Offset HetSO Homozygous Ratio Offset HomRO minor allele count nMinorAllele number of clusters Nclus number of AA calls n AA number of AB calls n AB number of BB calls n BB number of No Calls n NC and a hemizygous indicator hemizygous Values for five of these metrics CR FLD HetSO HomRO and nMinorAllele form the basis of the SNP classifications discussed below The CR FLD HetSO HomRO SNP QC metrics are described in Chapter 6 SVP OC Metrics SNP Metrics Used in the Ps Classification Step Step 8C Additional SNP QC tests used in the field are discussed in Additional SNP Metrics that may be Used for SNP Filtering on page 39 Step 8B Classify SNPs Using QC Metrics The Ps Classification function is used to sort the best probe set per SNP into seven classes based on five SNP QC metrics generated by the Ps Metrics function The classes are described in Figure 3 3 The seven classifications are based on default QC thresholds shown in Table 3 1 for different genome types Note the user can change the thresholds if desired The best probe set is determined by the classification priority order PolyHighRes NoMinorHom OTV MonoHighRes and CallRateBelowThreshold For a SNP with two probe sets where one probe set is NoMinorHom and one probe set is MonoHighRes the probe set that has been classified as NoMinorHom will be selected as the best probe set See the SNPolisher
44. apt probe set genotype script for step 7 bin apt probe set genotype log file lt OUTDIR gt apt probe set genotype log xml file lt ANALYSIS FILES DIR gt lt axiom_array gt _96orMore_Step2 r lt gt apt probe set genotype AxiomGT1 xml analysis files path lt ANALYSIS FILES DIR gt out dir lt OUTDIR gt summaries write models cc chp output cel files lt OUTDIR gt cel_list3 txt The generation of cel list3 txt is discussed in steps 5 and 6 Note that this example script for step 7 executes axiom array 960rMore Step2 r apt probe set genotype AxiomGT1 xml whereas the example script for step 4 executes axiom array 960rMore Stepl r apt probe set genotype AxiomGT1 xml The step 7 genotyping script includes options to write out a number of files to lt OUTDIR gt The default files are AxiomGT1 calls txt which contains the genotype calls coded into 0 1 2 and 1 for each probe set and sample a AxiomGT1 confidences txt which contains the confidence score described in What is a SNP Cluster Plot for AxiomGT1 Genotypes on page 10 for each genotype call in the AxiomGTI calls txt file a AxiomGTl report txt which contains information about each sample NOTE The output is per probe set not per SNP A probe set is a set of probe sequences interrogating a SNP site Although most SNP sites are interrogated by only one probe set and therefore there is usually a one to one
45. are colored gray o ui N N ej o N e 9 wo Q 4 2 0 2 4 6 4 2 O 2 4 6 Unusual Patterns of Relatedness Cross contamination of samples can cause samples to appear to be related to each other when examining their genotypes Depending on the extent of the cross contamination it can be just a pair of samples or entire sections of the plate that show increased relatedness Relatedness can be examined using the method described in the Relatedness section of Laurie et al Increased Computed Heterozygosity Cross contamination of samples will increase the computed heterozygosity relative to pure samples in the data set due to mixing of homozygous genotypes with heterozygous or opposite homozygous genotypes Note that poor quality pure samples will also exhibit increased computed heterozygosity The heterozygosity of a sample is the percentage of non missing genotype calls that are heterozygous AB The CHP summary table in GTC provides of AB calls for a sample under the het rate column het rate is displayed together with sample call rate call rate in the CHP summary table Additional Plate QC This section discusses general methods used in the field to detect outlier plates and batches It is not feasible to give absolute thresholds on most of these methods for outlier detection but careful consideration should be applied prior to including samples from flagged outlier plates in further analyses Evaluate
46. as more information on the fitTetra package Section 4 3 of the SNPolisher User Guide 1s a detailed example of how to run the functions in order to produce SNPolisher compatible calls confidences and posteriors files for auto tetraploid data Increase the Stringency for Making a Genotype Call Ps CallAdjust is a post processing SNPolisher function for rewriting less reliable SNP calls to No Call by decreasing Confidence Score thresholds Confidence Scores are discussed in What is a SNP Cluster Plot for AxiomGT1 Genotypes on page 10 A detailed description of Ps CallAdjust is given in Section 3 6 of the SNPolisher User Guide and examples of the effect of changing the threshold are described in Sections 4 1 5 and 4 2 7 in the SNPolisher User Guide 1 Voorrips RE Gort G Vosman B Genotype calling in tetraploid species from bi allelic marker data using mixture models BMC Bioinformatics 2011 May 19 12 172 ISSN 1471 2105 p 11 Additional Sample and Plate QC Additional Sample QC Detecting Sample Mix ups A critical component to a successful GWAS and other studies is that the identities of the samples in the study set are not confused during the sample and array processing For human samples Axiom arrays contain a set of Signature SNPs whose genotypes will uniquely identify the individual and GTC conveniently produces a signature SNP report in the pre genotyping QC process Affymetrix recommends checking that the number of
47. brary Files GTC APT lt axiom_array gt r lt gt ps2snp_map ps N A Required lt axiom_array gt _96orMore_Step1 r lt gt apt N A Required probeset genotype AxiomGT1 xml lt axiom_array gt _96orMore_Step2 r lt gt apt N A Required probeset genotype AxiomGT1 xml lt axiom_array gt r lt gt cdf Required Required lt axiom_array gt r lt gt qca Required Required lt axiom_array gt r lt gt qcc Required Required lt axiom_array gt r lt gt step1 ps Required Required lt axiom_array gt r lt gt genetic_prior txt Required Required lt axiom_array gt r lt gt AxiomGT1 sketch Required for human genomes Optional for non human genomes Required for human genomes Optional for non human genomes lt axiom_array gt r lt gt chrXprobes Required for mammalian genomes N A for non mammalian genomes Required for mammalian genomes N A for non mammalian genomes lt axiom_array gt r lt gt chrY probes Required for mammalian genomes N A for non mammalian genomes Required for mammalian genomes N A for non mammalian genomes lt axiom_array gt r lt gt specialSNPs Required for human genomes Required for non human genomes if gender calling is executed Required for human genomes Required for non human genomes if gender calling is executed probeset genotype AxiomGT1 xml lt axiom_array gt _LessThan96_Step1 r lt gt apt N A Required for small sample size probeset genotype AxiomGT1 xml lt
48. ch other Batches that appear to be outliers must be examined carefully to determine whether their deviation can be accounted for by biological characteristics of the samples which may be difficult in projects with multiple sources of ethnic variation and or relatedness among samples Pluzhnikov A Below JE Tikhomirov A Konkashbaev A Nicolae D Cox NJ 2008 Differential bias in genotype calls between plates due to the effect of a small number of lower DNA quality and or contaminated samples Genet Epidemiol 32 676 Laurie CC Doheny KF Mirel DB Pugh EW Bierut LJ Bhangale T Boehm F Caporaso NE Cornelis MC Edenberg HJ Gabriel SB Harris EL Hu FB Jacobs KB Kraft P Landi MT Lumley T Manolio TA McHugh C Painter I Paschall J Rice JP Rice KM Zheng X Weir BS GENEVA Investigators Quality control and quality assurance in genotypic data for genome wide associ ation studies Genet Epidemiol 2010 Sep 34 6 591 602 3 Pluzhnikov A Below JE Tikhomirov A Konkashbaev A Nicolae D Cox NJ 2008 Differential bias in genotype calls between plates due to the effect of a small number of lower DNA quality and or contaminated samples Genet Epidemiol 32 676 SNP QC Metrics SNP Metrics Used in the Ps_Classification Step Step 8C SNP Call Rate CR SNP Call Rate Samples Called N Samples Called the number of samples assigned a genotype call of either AA BB or AB at the SNP locus That is the number of samples that do not have a
49. de Both APT and GTC software tools require the files collectively referred to as analysis library files listed in Table 1 1 to appropriately process and interpret the data For Axiom arrays developed through the Axiom custom design program analysis files are made available from a secure file exchange server to the owner of the array The analysis files for Axiom catalog and expert arrays are available from either the array product page www affymetrix com or through direct download via GTC Table 1 1 lists the names of all analysis files used to process Axiom genotyping arrays in GTC or APT An annotation file is an additional file not required for genotyping and is not listed below but used in GTC to display SNP annotations in SNP results tables the cluster graph visualizations and for some export functionality Annotation files are available for download through GTC or on the array product page or the Secure File Exchange in the same locations as the analysis library files Table 1 1 Files Used For Analysis of Axiom Genotyping Arrays lt axiom_array gt will be replaced with the actual name of the array lt R gt will be replaced with the version of the analysis files For example when lt axiom_array gt Axiom_BioBank1 and lt R gt r2then lt axiom_array gt _96orMore_Step2 r lt gt apt probeset genotype AxiomGT1 xml Axiom BioBank1 96orMore Step2 r2 apt probeset genotype AxiomGT1 xml Analysis Li
50. ducing an FLD of 7 97 Figure 6 2 Examples of SNPs with low FLD A and high FLD B 39 2 D 1 2 3 SNP with low FLD 2 95 FLD detects that SNP with high FLD 7 97 The cluster the heterozygous cluster center and a centers are far apart from each other in homozygous cluster center are too close the X dimension together in the X dimension Chapter 6 SNP QC Metrics 37 Heterozygous Strength Offset HetSO SNP HetSO Het Strength Offset M M E ab bb A X bb M M aa bb Where M 4pp center of aa cluster etc Heterozygous strength offset HetSO measures how far the heterozygous cluster center sits above or below the homozygous cluster centers in the Y dimension Low HetSO values are produced either by mis clustering events or by the inclusion of samples that contain variations from the reference genome used to design the array probe Most well clustered diploid SNPs have positive HetSO values as shown in Figure 6 3 A HetSO of 0 39 Visually SNPs with low HetSO show average signal value along the y axis that is much lower for the heterozygous cluster than for the homozygous clusters Figure 6 3 B Figure 6 3 C Figure 6 3 D Figure 6 3 B shows a SNP with a very low HetSO value 0 82 This is an OTV SNP and should either be removed from the downstream genotyping analysis or be re analyzed with the OTV Caller function Figure 6 3 C shows a multi cluster SNP with one ver
51. e calls with GTC SNP Cluster Graph is also provided in this chapter a Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software provides instructions for executing the Best Practices steps 1 7 with APT combined with usage of SNPolisher functions for executing Step 8 Instructions for visualizing SNP cluster plots with SNPolisher function Ps Visualization is provided in this chapter 1 Manolio TA Collins FS The HapMap and genome wide association studies in diagnosis and therapy Annu Rev Med 2009 60 443 56 de Bakker PI Ferreira MA Jia X Neale BM Raychaudhuri S Voight BF Practical aspects of imputation driven meta analysis of genome wide association studies Hum Mol Genet 2008 Oct 15 17 R2 R122 8 3 Baker M Genomics The search for association Nature 2010 Oct 28 467 7319 1135 8 Laurie CC Doheny KF Mirel DB Pugh EW Bierut LJ Bhangale T Boehm F Caporaso NE Cornelis MC Edenberg HJ Gabriel SB Harris EL Hu FB Jacobs KB Kraft P Landi MT Lumley T Manolio TA McHugh C Painter I Paschall J Rice JP Rice KM Zheng X Weir BS GENEVA Investigators Quality control and quality assurance in genotypic data for genome wide associ ation studies Genet Epidemiol 2010 Sep 34 6 591 602 Background Axiom Array Terminology Marker Figure 2 1 Illustration of a Marker Forward strand probeset CAACATCATATGCAGGAGGTGCTCTGACAA G T JAMAA A ATAAGCAACTTGGATCTCCCCCAGT GTTGTAGTATACGTCCTCCACGA
52. e downloaded together into the same directory called lt ANALYSIS FILES DIR gt Example apt probe set genotype script for step 4 of the best practice analysis workflow using APT bin apt probe set genotype log file lt OUTDIR gt apt probe set genotype log xml file lt ANALYSIS FILES DIR gt lt axiom_array gt _96orMore_Step1 r lt gt apt probe set genotype AxiomGT1 xml V analysis files path lt ANALYSIS FILES DIR gt out dir lt OUTDIR gt cel files lt OUTDIR gt cel_list2 txt The generation of cel list2 txt is discussed in step 3 Note Choose lt axiom_array gt _LessThan96_Step1 r lt gt apt probe set genotype AxiomGT xml to perform QC genotyping with SNP specific models if batch size is less than 96 samples Best Practices Step 5 QC the Samples Based on QC Call Rate in APT Remove samples with a QC call rate value less than the default threshold of 97 To execute this filter step refer to the column call rate in the file lt OUTDIR gt AxiomGT1 report txt produced by step 4 When executing the workflow with APT GTC automates this step the user must write a script to remove CELs from the OUTDIR cel list2 txt whose call rate values are less than 97 We will refer to this CEL list as cel list3 txt Note that the AxiomGT1 report txt file will have a number of header lines beginning with The file can be read directly into a table data frame using the R read table function whic
53. e name and location of the AxiomGT1 posteriors txt and AxiomGTI calls txt file and the desired name and location for the new metrics output file Additionally Ps Metrics will calculate the metrics on only a subset of probe sets if a list of desired probe sets is supplied See the SNPolisher User Guide for a longer example of running the SNPolisher functions The input files required for Ps Metrics should either be located in the working directory or the user must provide the file path when typing the arguments If the file path is wrong or no file path is given and the files are not in the working directory R will return an error of file not found R working directories are discussed in the SNPolisher User Guide If a user has set the R working directory to the desired output folder location and the Step2_ AxiomGTI Step 7 Best Practices files are in the lt OUTDIR folder then the Ps Metrics command in R running on Windows would be gt Ps_Metrics posteriorFile lt OUTDIR gt AxiomGT 1 snp posteriors txt callFile lt OUTDIR gt AxiomGT1 calls txt output metricsFile metrics txt The output from Ps Metrics the default name is metrics txt is a text file containing the SNP QC metrics Each row is a SNP and each column is a QC metric The output should look similar to Figure 8 1 This output file will be one of the input files for other SNPolisher functions so the user must know the file s name and location on the computer
54. ecessarily mean that the performance is poor The most important metric for determining which plates should be included in the Best Practices Step 7 cluster set is the average QC call rate of passing samples Step 6 OC the Plates on page 16 Figure 5 3 Box Plots of DQC Values per Plate Outlier Time The solid green line near the top of aed represents the median of the medians across all plates The dashed lines represent 2 standard deviations from the median of medians The red line near the bottom of the graph at 0 82 indicates the recommended DQC threshold In this example the first plate identified by the arrow is an outlier because the upper bound of the box i e the 25th a the DQC mean of this sample plate is lower than 2 standard deviations below the median of medians Monitor Plate Controls As part of routine processing for large genotyping studies it is good practice to include at least one control sample with known genotypes on each plate e g a HapMap sample The calls obtained on the plate can be compared to the expected calls to obtain a measurement of genotyping concordance between the genotypes of the control samples and the genotypes of the known sample to help indicate whether there were plate processing or analysis issues A less robust but acceptable indicator of performance is to measure reproducibility by genotyping duplicate samples the genotypes of which may not be conclusively k
55. ecommendations are When plates have known significant differences for example when they have been processed at greatly different times many months apart or in different labs In these cases divide the plates into batches according to the date of processing and or the lab where the samples were run Users may attempt to co cluster plates with such differences but plate QC guidelines Step 6 OC the Plates on page 16 and Additional Plate OC on page 32 must be followed carefully DNA samples extracted from different tissues or with different techniques should be grouped into separate batches For example blood based saliva based and semen based samples should be grouped into separate batches DNA that is amplified with an extra WGA step should be grouped into a separate batch Polyploid samples with different genome ploidy levels should be grouped into separate batches Polyploid samples should not be genotyped together with diploid samples in a single batch Samples with auto polyploid and allo polyploid genomes should be grouped into separate batches Plant and animal samples from subpopulations that are greatly divergent from each other or from the array reference genome should be segregated and analyzed separately What comprises greatly divergent is a gray area and may require several rounds of Best Practices analysis to determine which subpopulations can be optimally batched together in a genotyping cluster run Methods for gen
56. ensities of probe sequences for non polymorphic genome locations i e sites that do not vary in sequence from one individual to the next When subject to the two color Axiom assay probes expected to ligate an A or T base referred to as AT non polymorphic probes produce specific signal in the AT channel and background signal in the GC channel The converse is true for probes expected to ligate a G or C base referred to as GC non polymorphic probes DQC is a measure of the resolution of the distributions of contrast values where AT Signal GC Signal Contrast AT Signal GC Signal Distributions of contrast values are computed separately for the AT non polymorphic probes which should produce positive contrast values and GC non polymorphic probes which should produce negative contrast values If sample quality is high then signal will be high in the expected channel and Chapter 3 Best Practices Genotyping Analysis Workflow 16 low in background channel and the two contrast distributions will be well resolved A DQC value of zero indicates no resolution between the distributions of AT and GC probe contrast values and the value of 1 indicates perfect resolution Step 3 QC the Samples Based on DQC Samples with a DQC value less than the default DQC threshold should be excluded from Step 4 Generate Sample QC Call Rate Using Step1 AxiomGTI These samples should be either reprocessed
57. er with substantially low hybridization intensities that are centered at zero in the X Contrast dimension and fall below the true AB cluster in the Y Size dimension OTV clusters are often miscalled as AB Figure 4 1 A The SNPolisher OTV Caller function performs post processing analysis to identify miscalled AB clustering and identify which samples should be in the OTV cluster and which samples should remain in the AA AB or BB clusters Samples in the OTV cluster are re labelled as OTV Figure 4 1 B SNPolisher OTV Caller intended usage is for SNPs that have been classified into the OTV class by the Ps Classification function Step 8B Classify SNPs Using QC Metrics on page 19 Instructions for executing OTV calling are provided in the SNPolisher User Guide see the SNPolisher User Guide for more details on OTV Caller Instructions for Generating SNP cluster plots for the recalled OTV genotypes and thus producing the 4th cluster colored cyan are provided in the SNPolisher User Guide see Ps Visualization Figure 4 1 Effect of OTV Calling on OTV cluster arrow genotypes A Before OTV genotyping the OTV cluster is mis called as AB gold B After OTV genotyping the OTV cluster has been identified and re labeled as a 4th OTV genotype cluster cyan e e S wo o o e o eo 3 2 1 0 1 2 3 9 2 1 OO 1 2 3 Didion JP Yang H Sheppard K Fu CP McMillan L de Villena FP Churchill GA Discovery of novel
58. esenting A and one representing B These two clusters should be easily resolved from one another and so are recommended by default Affymetrix still recommends that customers perform a visual check of the cluster plots to confirm this assumption The small number of SNPs from chromosome Y and the mitochondrial genome make it possible to visually inspect all of their SNP cluster graphs Chapter 3 Best Practices Genotyping Analysis Workflow 23 Figure 3 4 Cluster plots of mitochondrial and Y chromosome SNPs mitochondrial SNPs 80 85 90 95 wo 80 85 90 9 Panels A through F of Figure 3 4 show the expected pattern of homozygous genotype clusters for mitochondrial SNPs and panels G through L of 6 show the expected pattern of homozygous genotype clusters produced by the Y chromosome SNPs of male samples In Figure 3 4 Panel G L a cluster of No Call data is visible in addition to the one or two expected homozygous genotype clusters This No Call cluster is due to the presence of female samples within the data set Since female samples lack a Y chromosome these samples produce data points with a signal essentially equivalent to background signal that are automatically set to No Call in female samples NOTE it is important to exclude Y chromosome SNPs from the QC tests for X chromosome and autosomal SNPs because the inclusion of female samples in the data set will incorrectly cause Y chromosome SNPs to fail the
59. files were downloaded together in the same directory called ANALYSIS FILES DIR gt Example apt geno qc script for step 2 of the best practice analysis workflow bin apt geno qc analysis files path ANALYSIS FILES DIR aml file ANALYSIS FILES DIR gt lt axiom_array gt r lt gt apt geno qc AxiomQC1 xml cel files lt OUTDIR gt cel_list1 txt out dir lt OUTDIR gt out file lt OUTDIR gt apt geno qc txt log file lt OUTDIR gt apt geno qc log The generation of cel listl txt is discussed in step 1 Best Practices Step 3 Conduct Sample QC on DQC Remove samples with a DQC value less than the default DQC threshold of 0 82 To execute this filter step refer to the column axiom dishqc DQC in the file lt OUTDIR gt apt geno qc txt produced by step 2 of the best practice analysis workflow When executing the workflow with the APT system GTC automates this step the user must write a script to remove CELs from the OUTDIR cel listl txt with DQC values that are 0 82 We will refer to filtered CEL list from this step as cel list2 txt Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 63 Best Practices Step 4 Generate Sample QC Call Rates Using APT Genotype calls are produced by the program apt probe set genotype Below is an example script for a Linux command line shell In this and other example scripts with APT programs we assume that the analysis files wer
60. for Executing Best Practices Steps with GTC and APT Software 54 3 Enter a name for the SNP list and click OK The SNP list will be displayed in the data tree 4 Aftera SNP list has been imported right click a Genotype Results batch and select Show SNP Cluster Graphs from the shortcut menu that appears Figure 7 12 File Edt Workspace Window Tools Help VEU g um2277800U8dgom Waimes BE 42 CHF FT or 42 CHP Avon GW Hh j Show CHP Summary Table M Pe il Show SNP Summary Table GTC 42 CHP Data Se T Create SNP List gas bes Optimize for Cluster Graph Sample Attributes irtenety Data bes Show SNP Cluster Graphs Genctype Resuts Wy Creste Custom Data ek A2 EA intensky Data Group Copy Number LOH Resut L Export Genotype Results Copy Naber Vatation Fey Export Genotype Results for PLINK ES m d Export SNP Summary Table 42 CHP SNP dg Export Merged Genotype Results Run CHP vs CHP Concordance Check Run CHP vs TXT Concordance Check Rename Genotype Results Group Remove Genotype Results Group 5 Ifnone of the CHP files have matching sample files or if all of the CHP files have matching sample files no warning appears and the cluster graph is generated If some of the CHP files are missing matching sample files ARR a warning will appear that some CHP files do not have matching sample files Click Yes to produce a cluster plot with a gray spade for samples without t
61. gt e red yellow orange green blue purple 1 red yellow orange green blue purple The user cannot change the colors of the posterior and prior ovals yellow and light blue The default values for geno col are red gold blue gray cyan green darkgreen purple These colors correspond to the AA cluster the AB cluster the BB cluster null calls the OTV cluster highlighted samples a fourth cluster optional and a fifth cluster optional See Figure 8 3 for one SNP plotted with different colors Figure 8 3 One SNP Plotted With Color and Sample Highlighting Options A B ps101 ps101 AvM called genotypes AvM called genotypes 12 0 12 0 11 8 11 8 A B 2 11 6 A B 2 11 6 hi N N i 1 0 1 2 A B A B A One SNP plotted with default color options B The same SNP plotted with different and highlighted samples colors and the same highlighted samples The encoding for colors in R is quite complicated For more information on selecting colors in R see the SNPolisher User Guide In addition to setting colors through the geno col argument highlighted samples can be set to multiple colors using the text file given for sampFile see Figure 8 4 A In this case the text file is a two column tab deliminated file The first column is the sample name and the second column is the desired color The
62. h ignores lines beginning with 7 Best Practices Step 6 QC the Plates In this section we provide instructions for computing the basic plate QC metrics and guidelines for identifying plates to remove from the analysis Note that the user must write a script or use EXCEL to compute the plate QC metrics a Group the CEL files by plate then for each plate Compute plate pass rate a Plate Pass Rate Samples passing DQC and 97 call rate X 100 Total samples on the plate Compute the average call rate of passing samples on the plate Remove the samples that failed the sample QC tests in steps 3 and 5 Compute the average call rate of the remaining samples for the given plate a Guidelines for passing plates in average call rate of passing samples gt 98 5 If non passing plates are identified in step 6 then all samples from these plates must also be removed in the process of creating cel list3 txt Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 64 Best Practices Step 7 Genotype Passing Samples and Plates Using AxiomGT1 Step2 Step 7 produces genotype calls for all SNPs and passing samples Genotype calls are produced by the program apt probe set genotype Below is an example script for a Linux command line shell In this and other example scripts with APT programs we assume that the analysis files were downloaded together into the same directory called lt ANALYSIS_ FILES DIR gt Example
63. he attributes selected from the Color or Shape drop down lists 6 Next a SNP list dialog box appears Select a SNP list and click Okay If there are no SNPs in common between the SNP list and the array probes a warning will appear and no cluster plot will be generated If some but not all of the SNPs on the SNP list are in common with the array probes a warning notice will appear stating that some of the SNPs are invalid A new SNP list containing only the SNPs in common with the array probes will be created in the genotyping results output folder Import this SNP list to produce SNP cluster plots If the SNP list has no invalid SNPs the Select an Annotation File dialog box opens if an annotation file has not already been selected If an annotation file is not available on the computer the user will be prompted to download one The SNP cluster plot is displayed Figure 7 13 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 55 Figure 7 13 The SNP Cluster Plot Generated by the GTC SNP Cluster Graph Function Graph tab t Cluster Graph 20101008_101533_CHP i Select Files 2 Pry bg La X Color Genotype gt Shape Sex z AX 11089490 20101008_101533_CHP Axiom Chromosome 2 Graph tool bar 13 00 Genotype MAA E AB B BB B NoCall Sex OFemale OCO Male Oo Ellipses OBB QAB QAA e oB m o OA Prior 10 00 Be N Posterior 12 00 Cluster Graph of Strength 8
64. hresholds so HomRO flag should be set to FALSE Chapter 8 Instructions for Executing Best Practices Steps with Command Line Software 67 When the HomHet metric is set to TRUE default Ps_Classification will classify two cluster SNPs with one homozygote and one heterozygote cluster as NoMinorHom If set to FALSE SNPs will be classified as PolyHighResolution Missing the minor homozygote cluster is unreasonable for highly inbred species e g wheat This metric should be turned on when classifying probe sets in highly inbred species The two optional arguments that deal with conversion are output converted and priority order output converted is a logical indicator for outputting a list of converted recommended SNPs to the file converted ps default is FALSE priority order is used when performing probe set selection the best probe set is selected according to the priority order of probe set conversion types These are based on the default category order PolyHighResolution NoMinorHom OTV MonoHighResolution and CallRateBelowThreshold The priority order argument allows the user to change the order of categories when determining which probe sets are selected as the best probe set for a SNP All five of the listed categories must appear in priority order where the user specifies the order GTC is a flag indicating if the category files PolyHighResolution ps NoMinorHom ps Hemizygous ps MonoHighResolution ps CallRateBelowThreshold ps
65. ifferential bias in genotype calls between plates due to the effect of a small number of lower DNA quality and or contaminated samples Genet Epidemiol 32 676 Voorrips RE Gort G Vosman B Genotype calling in tetraploid species from bi allelic marker data using mixture models BMC Bioinformatics 2011 May 19 12 172 Zondervan KT Cardon LR Designing candidate gene and genome wide case control association studies Nat Protoc 2007 2 10 2492 501
66. iles for the array Table 1 1 on page 7 If this file has not been provided users should contact their local Affymetrix Field Application Support or send email to Support gaffymetrix com output dir Output directory for results files species type The type of organism genotyped The current options are Human non human Diploid and Polyploid output converted Flag indicating whether to output a list of converted SNPs to a file converted ps default is FALSE cr cutoff Threshold for call rate If not specified the default for human is 95 and for diploid and polyploid is 97 fld cutoff Threshold for FLD default 3 6 het_so_cutoff Threshold for HetSO default 0 1 het so otv cutoff Threshold for OTV detection default 0 3 hom ro 1 cutoff Threshold for HomRO when a SNP has one genotypes default 0 6 hom ro 2 cutoff Threshold for HomRO when a SNP has two genotypes default 0 3 hom ro 3 cutoff Threshold for HomRO when a SNP has three genotypes default 0 9 hom ro Flag indicating whether the metric HomRO is used in classification default is TRUE hom het Flag indicating whether the metric HomHet is used in classification The HomHet metric identifies two cluster SNPs probe sets with one homozygote cluster and one heterozygote cluster This checks if the minor homozygote cluster is missing which is unreasonable for highly inbred species e g wheat This metric should be turned on when classifyi
67. in the laboratory or dropped from the study The default DQC threshold value is 0 82 for all Axiom arrays except Axiom BOSI which is 0 95 Step 4 Generate Sample QC Call Rate Using Step1 AxiomGT1 Not all problematic samples are detectable by the DQC metric prior to the first round of genotyping see Detecting Mixed Contaminated DNA samples on page 30 To achieve the highest genotyping performance additional poor samples should be filtered post genotyping so that these samples do not pull down the cluster quality of the other samples The most basic post genotyping filter is based on the sample QC call rate For this step samples with passing DQC values are genotyped using a subset of probe sets usually 20 000 that are autosomal previously wet lab tested working probe sets with two array features per probe set If no probe sets on the array have been wet lab tested before array manufacturing this is the case for many arrays with non human SNPs Affymetrix requests the user to provide at least a plate of Axiom data to identify probe sets that meet this criteria Affymetrix will then provide the Axiom Analysis Library package Table 1 1 for the array Users should contact their local Affymetrix Field Application Support or send email to Support affymetrix com when such data is available This Best Practices Step 4 is referred to as Step1 AxiomGTI genotyping in the instructions provided for genotyping with GTC Chapter 7 and APT Chapter 8
68. including SNP call rate and HWE values can be viewed in the SNP summary table Prior to viewing the SNP summary table in GTC create a SNP list 1 e a list ofthe SNPs of interest that have been retained after SNP QC filtering using the Create SNP List menu option in GTC Next select Show SNP Summary Table from the same menu to display the SNP call rate and other summary metrics Mendelian Trio Error Mendelian errors can be detected in parent offspring trios Mendelian trio error rate is calculated as the number of errors detected in a particular family divided by the number of families in which the offspring and parents have available genotypes This method of error detection is less efficient than other methods because many genotyping errors are consistent with Mendelian inheritance e g the offspring of AB and BB parents may have a true BB genotype but is called as AB and this error will not influence the Mendelian trio error rate SNPs that have high Mendelian error rates in the study should be examined in cluster plots for symptoms of mis clustering Genotyping Call Reproducibility SNP genotyping error rates can be estimated from the reproducibility of genotype calls excluding No Calls of replicated samples One approach is to use duplicated pairs of samples and count the number of pairs with discordant calls Given that mean error rates are low a large number of duplicated pairs is required to provide enough precision to meaningfull
69. iors are solid for both the SNPolisher and GTC cluster plots Unless specified otherwise all cluster plots in the document have been produced by Ps Visualization and use the sample colors and shapes as described above Figure 2 2 SNP Cluster Plot produced by the SNPolisher package via the Ps Visualization function uw YT Blue BB Gold AB Red AA Grey NoCall m Solid Ellipses Size centered around Posterior wo cluster locations c P Dashed Ellipses centered around Prior cluster locations 10 0 Contrast X Contrast LogRatio Log2 A signal B signal Y Size Strength Log2 A signal Log2 B signal 2 Chapter 2 Background AX 11086525 20110121 162355 CHP Axiom HapMap 2 12 31 11 86 10 96 10 52 4 40 2 26 2 2 4 17 Log Ratio 12 Best Practices Genotyping Analysis Workflow Design the Study to Avoid Experimental Artifacts Good experimental design practices include randomizing as many processing variables as possible For a GWAS this means distributing the cases and controls across sample plates not processing all samples of one type on one day or having one individual or laboratory process the controls and another process the cases For larger studies it is suggested that the experimental design include at least one control sample of known genotype on each plate e g a HapMap sample to serve as a processing control The geno
70. ld blue gray cyan green darkgreen purple nclus 3 In this example the working directory contains the output from Ps Classification including the PolyHighResolution SNP list The output PDF file will be named Cluster PolyHighResolution pdf and will be made in the working directory There is no reference genotype file The prior distributions will be plotted but no prior information file is given so it will be a generic prior The maximum number of SNPs plotted will be 6 The genotype colors given are the default colors and there are three clusters per SNP See SNPolisher User Guide for a longer example of running the SNPolisher functions and for more details of Ps Visualization References Affymetrix Genotyping Console User Manual Genotyping Console 4 2 supported on Windows 7 64 bit and Windows 8 1 http media affymetrix com support downloads manuals gtc 4 2 user manual pdf a Genotyping Console 4 1 supported on 32 bit and or Windows XP systems http media affymetrix com support downloads manuals gtc 4 1 user manual pdf Affymetrix Power Tools Manual Manual apt probeset genotype 1 16 1 http media affymetrix com support developer powertools changelog apt probeset genotype html Affymetrix 2007 BRLMM P a genotype calling method for the SNP 5 0 array Technical Report Baker M Genomics The search for association Nature 2010 Oct 28 467 7319 1135 8 Cardon LR Palmer LJ Population stratifica
71. ld be reviewed for plate processing problems If there are no known plate processing problems the user may proceed with caution to include passing samples from such plates Low sample pass rates may be caused by problematic sample sources for some but not all of the samples As long as such samples are excluded by steps 2 5 the remaining samples may be included All samples on non passing plates red zone Figure 3 2 should be excluded from the Best Practices step 7 genotyping run and samples on such plates should be reprocessed The occurrence of non passing plates should be rare 5 If the occurrence is higher the lab is recommended to review the sample sources and or plate processing practices with the local Affymetrix Field Application Support person Chapter 3 Best Practices Genotyping Analysis Workflow 18 Figure 3 2 Graph of plate metrics for a batch of 39 plates of blood derived samples Each plate is shown as a black dot The graph is divided into three quality zones The gray and green zones with Mean QC call rates of passing samples 2 98 5 are the zones for passing plates The green zone flags high quality plates with gt 95 sample pass rate for the plate vertical red line on the right hand side of the graph and the mean sample QC call rate of passing samples 9996 per plate horizontal red line The gray zone flags marginal plates that should be subject to further review The red zone flags non passing plates that should
72. luster Graph 52 To Importa SNP LISt kb etiidi RR RR EAR RS bees dde RPM stad 53 Display a Particular SNP aao sace Roe edt ed eal doa RC ode td 55 Select a Single Sample o n nannan 56 Chapter 8 Appendix A Contents 5 Select Multiple Samples 0 a 57 Manually Change a Sample s Call 2 2 0 0 00 00000 cece 57 Lasso FUNCUOMN ie dai t eea pp ee SBR ne ee Re Ua E MEER RES 58 Saving a Cluster Plot 2 2 eee eee 59 Export Genotypes with GTC Software 0 aaas saaara e 61 Instructions for Executing Best Practices Steps with Command Line SOWA as oru cnc Gn EB UP AERE GAP OE Rt d NE NA Eq 62 Execute Best Practice Steps 1 7 with APT Software 0 0 0 0 0c ee 62 Best Practices Step 1 Group Samples into Batches 00 0000 2 cee ee eee 62 Best Practices Step 2 Generate the Sample DQC Values Using APT 62 Best Practices Step 3 Conduct Sample QC on DQC 2 lisse 62 Best Practices Step 4 Generate Sample QC Call Rates Using APT 0 63 Best Practices Step 5 QC the Samples Based on QC Call Rate in APT 63 Best Practices Step 6 QC the Plates 2 see 63 Best Practices Step 7 Genotype Passing Samples and Plates Using AxiomGT1 Step2 64 Execute Best Practice Step 8 with SNPolisher Functions a asuaan naana a aaraa 65 ire EU seeen o i a a arai a a a e ORA OT 65 Best Practices Step 8A Run Ps Metrics 20 65 Best Practices Step 8B Run Ps Classification
73. n study objective is a GWAS for a diploid genome Use of generic priors for large batches allows the genotyping algorithm to dynamically adapt to observed cluster locations and tends to maximize the number of well clustered SNPs in a given batch For small sample sets SNP specific priors are used to help the genotyping algorithm accurately call genotypes in Chapter 3 Best Practices Genotyping Analysis Workflow 19 the absence of observed intensities for the minor allele All Axiom arrays are provided with analysis files Table 1 1 on page 7 for genotyping large batches and some arrays are provided with analysis files for genotyping small batches Certain arrays may benefit from usage of SNP specific priors even when the sample size is large These may include arrays for genomes with large SNP specific variation in cluster locations such as such as allo polyploid genomes discussed below arrays with a large fraction of SNPs that are monomorphic in the population and arrays whose intended usage is genomic selection Creation and testing for the appropriate SNP specific priors requires study specific development NOTE The Best Practices Step 4 Sample QC call rates Step1 AxiomGT1 often run higher than the Sample call rates produced in Best Practices Step 7 Step2 AxiomGT1 This is because only tested working SNPs are used for Step 4 QC call rates whereas Step 7 call rates are often computed over untested probe sets with unpredictable perfo
74. nd for all of the cluster plots The remaining pages display six graphs per page Figure 7 20 Genotyping Console Example AX 11123387 AX ATI2338 cc chp Poly WighRizschtion_rand200 ue 1126968 AX A120M8 cech Polyt gsRaschuon rand200 ua 1173451 AL IN1T2481 ccchp PalyHtignRewshison_rand200 AX 11124020 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software Genotyping Console Example AX 11214207 LE AX AZHOU iccchg PetyignRizeolition_rand200 AX Affymetnic AX 11215377 AL 11218377 ce ch PohyetgRenciuton ran 209 ME fe AX 11244824 AX M2UAEM cc chp Pohyighsenciuton rand200 a T AX 11335446 ne 6 30 2014 6 30 2014 60 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 61 Export Genotypes with GTC Software The genotype calls for passing samples and recommended SNPs can be exported from GTC for downstream analysis with third party software One option is to directly export genotype calls to text files through GTC Figure 7 21 This export can be limited to a desired SNP list that has already been imported into GTC Once output text file s are generated downstream analysis can be performed using third party software It is also possible to export genotype calls in the P
75. ng SNP cluster plots are provided by two Axiom software systems 1 the SNPolisher package via the Ps Visualization function example shown in Figure 2 2 and 2 GTC via the SNP Cluster Graph function example shown in Figure 2 3 Instructions for the Cluster Graph and Ps Visualization function usages are provided in Chapter 7 and Chapter 2 respectively AxiomGT1 is a tuned version of the BRLMM P clustering algorithm that adapts pre positioned genotype cluster locations called priors to the sample data in a Bayesian step and computes three posterior cluster locations Genotype cluster locations are defined by 2D means and variances for AA AB and BB 1 Affymetrix 2007 BRLMM P a genotype calling method for the SNP 5 0 array Technical Report Chapter 2 Background 11 genotype cluster distributions Priors can be generic meaning the same pre positioned location is provided for every SNP or SNP specific meaning the different pre positioned locations are provided on a SNP by SNP basis AxiomGT1 clustering is carried out in two dimensions dimension Y is calculated as Log2 A_ signal Log2 B_signal 2 and dimension X is calculated as Log2 A signal B signal X carries the main information for distinguishing genotype clusters The X dimension is called Contrast in cluster plots produced by SNPolisher and og ratio in cluster plots produced by GTC The Y dimension is called Size in cluster plots produced SNPolisher and Strength in clu
76. ng SNPs probe sets in highly inbred species default is TRUE num minor allele cutoff Threshold for the number of minor alleles default 2 priority order When performing probe set selection the best probe set is selected according to the priority order of probe set conversion types Valid values are sequences of the strings PolyHighResolution NoMinorHom OTV MonoHighResolution and CallRateBelowThreshold separated by commas with each string occurring exactly once default is PolyHighResolution NoMinorHom OTV MonoHighResolution CallRateBelow Threshold Ps Classification reads in the SNP QC metrics table usually named metrics txt generated by Ps Metrics and performs SNP classification based on the customizable criteria It classifies SNPs probe sets into seven major categories PolyHighResolution NoMinorHom MonoHighResolution OTV CallRateBelowThreshold Other and Hemizygous The metric hom ro should be turned off for organisms where the homozygote clusters may center at zero in the contrast space particularly polyploid species If a ps2snp file is provided then the best performance probe set will be selected to represent SNPs with multiple probe sets There are seven output files one for each category containing lists of probe set IDs belonging to each category There is also a summary output file Visualize SNPs and Change Calls if Desired through GTC Plotted Cluster Graph
77. ng with genetic OTV_Caller function knowledge MonoHighRes SNPs are recommended with caution especially if the best probe set for the SNP site has never been tested An additional test for recommending MonoHighRes SNPs is to require that both probe sets if available on the array for the SNP site are classified as MonoHighRes and that the genotypes agree Hemizygous SNPs are recommended by default but visually inspection is advised see below SNPs that are classified as OTV may also be considered converted after the OTV Caller function has been used to re label the genotype calls see Adjust Genotype Calls for OTV SNPs on page 28 and after visual inspection of the recalled genotypes A total list of unique probe sets for recommended SNPs can be created manually by combining the category files described above for the default recommended yellow and or chosen by the user Or Ps Classification can be executed with output converted TRUE the default is FALSE and PolyHighRes NoMinorHom MonoHighRes and Hemizygous classes are combined to create a category file called converted ps Therefore converted ps contains all probe sets one per SNP that are recommended for downstream analysis if the Genome Species Type is human or Diploid outbred or mixture of inbred and outbred Visual SNP Analysis for Hemizygous SNPs Chromosome Y W and mitochondrial and other hemizygous genomes produce only two genotype clusters i e one repr
78. nown as they are with HapMap samples and then comparing the genotype reproducibility measurement between the duplicated samples In addition the gender call for each replicate of the sample should be the same As with the DQC plots the concordance value of the controls at the plate level should be tracked over time to detect trends and or outlier plates Chapter 5 Additional Sample and PlateQC 34 Check for Platewise MAF Differences Assuming a randomized study design the SNP minor allele frequency MAF values ona given plate should not systematically differ from the MAF values for the same SNPs on the remainder of the plates Such a shift in MAFs may reflect mis clustering events over the samples on such plates A chi squared analysis is a simple method for automatically detecting this type of effect Pluzhnikov et al 2008 A description of this method as described in Laurie et al 2010 and summarized here To detect batch effects on allelic frequencies we use a homogeneity test suggested by N J Cox Pluzhnikov et al 2008 3 If p is the sample minor allele frequency for a SNP on the i th plate with n samples p is the average frequency over all plates except the i th a total of i samples and pis the average over all plates a total of samples then a 1 degree of freedom chi squared test statistic is given by no Po for each SNP These statistics are averaged across SNPs to measure how different the plates are from ea
79. nsity Data In Bound Copy Number Copy Number Show Intensity QC Table Show Signature Genotypes Copy Number LOH Analysis Perform Copy Number Variatior Rename Intensity Data Group X Remove Intensity Data Group After the QC genotyping has been completed in GTC identify outlier samples with a sample call rate lower than 97 in the CHP summary table Next create an intensity data group that excludes the CEL files for these outlier samples After the sample level QC is done it is recommended to perform plate level QC for each plate to calculate the following metrics using Table 7 1 Table 7 1 Metric Calculation Minimum Criteria Average call rate of passing samples on the plate Plate pass rate a 100 None Total samples on the plate Average call rate of passing AVERAGE call rates of samples passing DQC and 97 call rate 2 98 5 samples on the plate NOTE The table gives the minimum criteria for a passing plate please see Step 6 QC the Plates on page 16 for more discussion about these plate metrics Also see Additional Plate QC on page 32 These metrics can be manually calculated in Excel or with a script using data from the CHP summary table Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 47 Step 7 Genotype all Passing Samples and Plates from the Same Batch Using All SNPs The purpose of this step is to perform genot
80. og in using your email address and click DevNet Tools SNPolisher is available under the Analysis Tools menu Download the zipped SNPolisher folder SNPolisher package zip The zipped folder contains the R package file SNPolisher XXXX tar gz where XXXX is the release number the User Guide the quick reference card the help manual the license copyright and readme files a PDF with colors for use in R and the example R code and four folders with example data for running in R Note that this zipped folder is not a package binary for installing in R Users must unzip the file to extract the SNPolisher folder which contains the tar gz package file Follow the instructions in the SNPolisher User Guide for installation instructions for R Perl and SNPolisher This section assumes some familiarity with the R programming language Please see the SNPolisher User Guide for instruction on R basics or contact your local FAS for support or send email to Support affymetrix com NOTE Step 8 also be can be executed using APT version 1 16 1 or higher using the Ps Metrics and Ps Classification software Best Practices Step 8A Run Ps Metrics After perl and R have been installed and the SNPolisher library has been loaded see SNPolisher User Guide the Ps Metrics function can be run Ps Metrics uses two output files from Best Practices Step 7 AxiomGT1 Step2 above as inputs AxiomGTI posteriors txt and AxiomGT1 calls txt The user should specify th
81. olisher functions requires the user to have some familiarity with the programming language R The R package files to install SNPolisher are available on the Affymetrix website www affymetrix com Select Register at the top of the website to register your email address with Affymetrix From the Partners and Programs menu select Developers Network Click DevNet Tools on the left side of the menu SNPolisher is available under the Analysis Tools tab Download the zipped SNPolisher folder Genotyping Console 4 2 User Manual http media affymetrix com support downloads manuals gtc_4_2_user_manual pdf 2 Affymetrix Power Tools User Manual http media affymetrix com support developer powertools changelog apt probe set genotype html Chapter 1 Introduction to Axiom Data Analysis 7 SNPolisher_package zip The zipped folder contains the R package file SNPolisher XXXX tar gz where XXXX is the release number the user guide the quick reference card the help manual the license copyright and readme files a PDF with colors for use in R and the example R code and two folders with example data for running in R Note that this zipped folder is not a package binary for installing in R Users must unzip the file to extract the SNPolisher folder which contains the tar gz package file For instructions on R basics installation and usage of the R functions including additional function not discussed in this document see the SNPolisher User Gui
82. onfiguration from the drop down menu to perform QC genotyping with generic priors model file if the batch size is equal or more than 96 samples Figure 7 4 n Choose the LessThan96 Stepl AxiomGTI as the analysis configuration from the drop down menu to perform QC genotyping with SNP specific models file if the batch size is less than 96 samples Figure 7 4 Perform genotyping from the GTC tool bar menu Select Perform Genotyping icon from tool bar Select a user intensity data group Analysis Configuration Parameters Algorithm S6orMore_Step1_AxiomGT1 Confidence Threshold 0 15 LJ JJ CJ None Hints File 9 Select Output Root Path C data LAT_CEL m Select Base Batch Name 20111123 150045 CHP Output File Suffix leave blank for no suffix v Use Defauit Base Batch Name as Posterior Models File Name Select Posterior Models File Name posterior models 20111123 150045 CHP Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 45 a Select the Intensity Data option from the Workspace drop down menu and then select Perform Genotyping select In Bounds and click OK Figure 7 5 o Choose the 96orMore Stepl AxiomGTI1 as the analysis configuration from the drop down menu if the batch size is equal to or more than 96 samples Refer to Figure 7 4 for more details n Choose the LessThan96_Step1_AxiomGT1 as the analysis configuration from the drop down menu to perform
83. otyping divergent subpopulations require exploration by the user One approach is to co cluster the divergent populations and attempt to identify a subset of working SNPs for the population spectrum Another approach is co cluster only samples from the separated divergent population identify a sub population set of working SNPs Our guideline for maximum batch size is 50 Axiom 96 Array Plates per batch This is based on internal Affymetrix analysis on the effects of batch size on genotyping quality as well as achieving reasonable computation performance of the command line analysis programs APT and SNPolisher see Chapter 8 with the system that will analyze the array plate batches As a reference point a batch size of 55 Axiom 96 Array Plates each with 650K probe sets requires about 16 hours to execute step 7 Figure 3 1 using the apt probe set genotype command Best Practices Step 7 Genotype Passing Samples and Plates Using AxiomGTI Step2 on page 64 on a Linux server with the following configuration x86_64 architecture 16 x 3GHz XEON core and 128 GB of RAM Note that this is without any computational parallelization Step 2 Generate Sample DQC Values Before performing genotyping analysis on any samples the quality of each individual sample should be determined Steps 2 through 5 collectively identify poor quality samples using first a single sample metric Dish QC DQC followed by sample QC call rate test DQC is based on int
84. otyping Analysis Workflow 22 The Ps_Classification function also selects the best probe sets from the Ps performance txt file and divides these into seven category files named PolyHighResolution ps NoMinorHom ps Hemizygous ps MonoHighResolution ps CallRateBelowThreshold ps Other ps and OTV ps Each category file is a tab delimited text file with probeset_ids for the category Each file has a column header called probeset_id Note that ps extension is an Affymetrix convention to indicate the file contains a list of probe set IDs Step 8C Create a Recommended SNP List SNPs that are not sorted into recommended classes for the genome type should be excluded from further downstream analysis Table 3 3 shows which classes are recommended for the given genome type SNPs in recommended classes are also referred to as converted in this document Table 3 3 Recommended SNP Classes Based on Genome Type and SNP Class Figure 3 3 Genome Type SNP Class determined by Ps Classification Function PolyHighRes NoMinorHom MonoHighRes Hemizygous OTV Human Recommended after genotyping with OTV_Caller function Diploid inbred Recommended after genotyping with OTV_Caller function Not Recommended only Diploid outbred or mixture of inbred and Recommended after genotyping with outbred OTV_Caller function Polyploid Requires Not Recommended after additional Recommended genotypi
85. populations Users should contact their local Affymetrix Field Application Support or send email to Support affymetrix com for usage of the inbred penalty option Figure 3 8 Cluster Plots for Inbred Populations A Allo polyploid plant B and C Diploid Plants q N z A M o o N oc A N A e i i o Y A q o o N 2 4 0 1 2 3 2 4 0 1 2 3 2 4 0 1 2 A Allo polyploid Plant B Diploid Plant C Diploid Plant Additional Genotyping Methods Manually Change Genotypes In some cases SNPs called incorrectly due to problematic cluster patterns can be corrected with expert manual intervention cases include SNPs with cluster splits and some cases of multi cluster SNPs such as OTV cluster patterns which escape the OTV classification Instructions are provided in Visualize SNPs and Change Calls if Desired through GTC Plotted Cluster Graph on page 52 Adjust Genotype Calls for OTV SNPs One of the SNP categories produced by the Ps_Classification function is OTV The term off target variant OTV are SNP sites Didion et al 2012 whose sequences are significantly different from the sequences of the hybridization probes for some or all of the samples in the batch OTV sites have reproducible and previously uncharacterized variation that interfere with genotyping ofthe targeted SNP Interference may be caused by double deletions sequence non homology or DNA secondary structures OTV SNPs display an OTV clust
86. r executing steps 8 of the best practice analysis workflow using Ps_Metrics and Ps_Classification programs from APT Version 1 16 1 or higher combined with some simple scripts to be written by the user See Affymetrix Power Tools information for download instructions Note that the same Step 8 results are produced when executing SNPolisher functions Ps_Metrics and Ps_Classification If SNPolisher usage is preferred please go to Execute Best Practice Step 8 with SNPolisher Functions on page 65 The example scripts below are directly usable by Linux users For readability the Linux commands are broken into several lines with the backslash character The backslash character is not recognized by the Windows OS The APT commands can also be executed on a Windows computer To execute the commands scripts Windows users should 1 Remove the backslashes and put the given command on one line 2 Change the forward slash to a backslash V when the input is a directory path 3 Enter the command in the Windows command prompt window Locate GTC Step2_AxiomGT1 Step 7 Output Files The files used to run the Step 8A Ps Metric software are generated from running Step2 AxiomGTI Step 7 Genotype all Passing Samples and Plates from the Same Batch Using All SNPs on page 47 genotyping in GTC as discussed above also see chapter 7 Genotyping Analysis in the Genotyping Console User Guide P N 702982 The user must take care to use
87. r graph The selected sample will be checked in the Sample Table and the Checked column in the table will display a check mark Figure 7 14 56 Figure 7 14 Selecting a Single Sample BS SNP Summary 20110121_162355_CHP LJ SNP List Axiom HapMap 2 bts Cluster Graph 20110121_162355_CHP i M Color Genotype Shape Genotype Select Files gt d b i AX 11086525 20110121 162355 CHP Axiom HapMap 2 12 31 Genotype QAA Was ABB 4 HoCall Ellipses Strength a EE 2 10 96 Prior N Posterior v Show Prior Models Show Posterior Models SNP Summary Table Sample Table Default View Log Checked File Differ ac 25 60749 74 08237 0 958 3 87 25 26789 7451111 0 931 364 2548216 74 16769 0 313 346 Yy gt computed gender call_rate hetrate hom_rate Pa NA18948 Axiom Gw Hu SNP A AxiomGT1 chp N 18949 Axiom Gw Hu SNP A AxiomGT 1 chp NA 18351 Axiom Gw Hu SNP A AxiomGT1 chp Jii Rows 270 Cols 14 Selected Rows 0 male 99 68987 99 77899 99 64985 female female The mouse is hovering over the selected sample The cluster plot shows the name of the CHP file for the sample The corresponding row of the Sample Table has a check mark in the Checked column for the sample Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 57 Select Multiple Samples To select multiple samples draw a closed shape around a gro
88. r more information Figure 3 7 Cluster Plot for an allo octoploid plant Each sample is colored by the AxiomGT1 genotype call blue gold red for the sub genome with the segregating allele Each genotype cluster is labeled by the likely allo octoploid genotype using the following notation the genotypes of 4 sub genomes are separated by dashes the genotype of the sub genome with the segregating allele is noted first red followed by the genotypes of the sub genomes whose alleles are fixed black It is likely that the genotypes of the fixed sub genomes are AA because clusters are shifted to the right in Contrast space which occurs when the A genotype dosage is higher than the B dosage eo BB AA AA AA t L o AB AA AA AA eo Led AA AA AA AA e o Chapter 3 Best Practices Genotyping Analysis Workflow 27 SNP Cluster Plot Patterns for Inbred Populations Inbred populations produce few or no heterozygous genotypes and there is often a high frequency of both of the homozygous genotypes Figure 3 8 All cases will be classified as PolyHighResolution as long as the Polyploid or Diploid option is selected in the Ps_Classification step However the SNP producing cluster plot Figure 3 8 C with no Heterozygous genotypes will be classified as Other if the Human option is selected in the Ps Classification step AxiomGT1 analysis options should be set to include the inbred penalty when genotyping inbred
89. rm Copy Number Variation Analysis Rename Intensity Data Group Remove Intensity Data Group Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 43 a Right click the All data group displayed in the data tree on the left and select the Perform QC option from the popup menu that appears Figure 7 3 B Workspace Axiom_GW_LAT iB Data Sets S Axiom_GW_LAT Axom KP UCSF LAT E Sample Attributes S i Intensity Data Rename Intensity Data Group Remove Intensity Data Group After the QC process has been completed the resulting DQC values and other quality metrics are displayed by GTC in an Intensity QC table Any samples with a DQC value 0 82 are highlighted and grouped into the Out of Bounds group to be excluded from the genotyping analysis For further information about the QC thresholds and the other metrics please refer to the Genotyping Console User Guide P N 702982 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 44 Steps 4 5 and 6 Generate QC Sample Call Rates QC the Samples and QC the Plates Perform QC genotyping for the group of CEL files that have passed the DQC threshold DQC gt 0 82 i e those CEL files in the In Bounds data group by performing any one of the following operations Select the Genotyping icon 3 on the tool bar menu Figure 7 4 o Choose the 96orMore Stepl AxiomGTI1 as the analysis c
90. rmance Step 8 Execute SNP QC The purpose of Step 8 is to identify probe sets that produce well clustered intensities see Evaluate SNP Cluster Plots on page 23 and whose genotypes are recommended for statistical tests in the downstream study When more than one probe set has been designed to interrogate a SNP the best probe set will be identified The overall approach is to sort the best probe set per SNP into categories based on a set of SNP QC metrics and then create a recommended probe set list for the downstream analysis The options for categorizing SNPs are based on thresholds for the SNP QC metrics where some thresholds have been adjusted for certain types of genomes Steps 8 uses the Ps_Metrics and Ps_Classification functions These functions are available in the SNPolisher R package and APT software version 1 16 1 or greater Instructions for SNPolisher R package usage are provided in Execute Best Practice Step 8 with SNPolisher Functions on page 65 and for APT usage in Execute Step 8 with APT Version 1 16 1 or Higher on page 49 The Ps_Metrics and Ps_Classification functions are not currently available with the GTC software Step 8A Create SNP QC Metrics The Ps_Metrics function is used on the output files from the Best Practices Step 7 genotyping run also referred to Step2 AxiomGT1 and computes twelve SNP QC metrics for each probe set probeset_id that was genotyped in Step 7 Call Rate CR Fisher s Linear Discrimin
91. s that includes SNPs that are interrogated with more than one probe set This file lt axiom_array gt r lt gt ps2snp_map ps should be provided with the Analysis Library Files for the array Table 1 1 on page 7 If this file has not been provided users should contact their local Affymetrix Field Application Support or send email to Support affymetrix com Below is an R command example for Ps Classification which 1 uses output from Ps Metrics metrics results in metric txt 2 the classification results should be stored in the folder called Output 3 the genotype data is human 4 ps2snp txt file ANALYSIS FILES DIR Axiom BioBankl r2 ps2snp map ps ANALYSIS FILES DIR means the full path to the Analysis Library file directory gt Ps Classification metricsFile metrics txt output dir Output SpeciesType Human ps2snpFile ANALYSIS FILES DIR Axiom BioBankl r2 ps2snp map ps Eight of the optional 15 arguments are classification thresholds for the QC metrics If only a species type is given Ps Classification will use the default thresholds for that genome type see Table 3 1 on page 21 SNPs classified as PolyHighRes must have SNP QC values that pass all of the thresholds There are two logical indicators HomRO flag indicates if HomRO thresholds should be used default is TRUE and HomHet flag indicates if the HomHet metric should be used default is TRUE Polyploid genotypes do not use either of the HomRO t
92. set for SNP in a diploid genome with well clustered intensities left and an example of a probe set with mis clustered intensities right A well clustered diploid genome SNP should have one to three approximately elliptical clusters with the center of each cluster reasonably separated from the centers of the other clusters and the position of the heterozygous cluster equal to or higher than the position of the homozygous clusters The mis clustered SNP example right is an example of cluster split where the correct BB genotype cluster has been incorrectly split into two clusters BB and AB and the some of the BB samples are incorrectly called AB gold In addition the correct AB cluster has been mislabeled as an incorrect AA cluster red The miscalled AB cluster is lower on the Y axis than the BB cluster This mis clustering event is easily detected by the SNP QC metrics CallRate HetSO and FLD and should be classified into the Other category Genotype calls for such SNPs may be manually recalled using the SNP Cluster Graph function in GTC Chapter 7 Figure 3 5 Examples of a well clustered SNP left and misclustered SNP right in Contrast vs Size space gp E BB a o e Cam o Q e 97 E 77 AA e Lr i AB N e e oa T T T T T 4 2 0 2 4 A B Multi cluster SNP Cluster Plot Patterns When a subset of samples in the batch co cluster in their own intensity space more th
93. st above the threshold See Table 3 1 on page 21 for default CR thresholds used in the Ps Classification step Chapter 6 SNP QC Metrics 36 Fisher s Linear Discriminant FLD IM ab n M Fisher s Linear Discriminant FLD Min i aa bb LU s Where M center of het cluster in log ratio dimension M M pp 7 center of hom a b cluster in log ratio dimension sd square root of variance pooled across all three distributions FLD is undefined when there is only one genotype cluster FLD is a measurement of the cluster quality of a SNP High quality SNP clusters have well separated centers and the clusters are narrow High quality clusters can be identified by examining the shape and separation of the SNP posteriors that are produced during genotyping FLD is essentially the smallest distance between the heterozygous middle cluster center and the two homozygous cluster centers in the X dimension CR and FLD are generally correlated but in some cases FLD will detect problems that are not captured by CR HomFLD is a version of FLD computed for the homozygous genotype clusters HomFLD is undefined for SNPs without two homozygous clusters Figure 6 2 A shows an example of a SNP with low FLD In this case the clustering algorithm has found the location of the BB cluster to be too close to the AB cluster producing an FLD of 2 95 In contrast the well clustered SNP in Figure 6 2 B has a high CR and separated cluster centers pro
94. ster center should be located approximately at 0 on the X axis If the clusters are shifted from their expected positions then the heterozygous clusters will be far away from zero A negative or low value of HomRO generally indicates that the algorithm has mislabeled the clusters The AA cluster should be on the right side of zero positive Contrast values and the BB cluster should be on the left side of zero negative Contrast values A negative HomRO value implies that one of the homozygous clusters is on the wrong side of zero Figure 6 4 A shows a misclustered SNP with a negative HomRO value 0 51 The homozygous BB cluster blue is on the wrong positive side on the x axis and the heterozygous AB cluster yellow is not over zero on the x axis Figure 6 4 B shows a well clustered SNP with a positive HomRO value 2 21 where the AA red cluster is to the right of zero the AB cluster yellow is over zero and the BB cluster blue is to the left of zero as expected Figure 6 4 Examples of SNPs with low HomRO A and high HomRO B wo o e o e ET S 9 3 E ET o o e e e eo eo o e 4 2 0 2 4 4 2 0 2 4 A B A B SNP with low HomRO 0 51 SNP with high HomRO 2 21 Chapter 6 SNP QC Metrics 39 Additional SNP Metrics that may be Used for SNP Filtering This section describes additional SNP metrics Hardy Weinberg p value Mendelian trio error and Genotyping call Reproducibility that
95. ster plots produced GTC AxiomGT1 genotype calls are made by identifying the genotype intensity distribution AA AB or BB each sample is most likely to belong to The samples are colored and shaped by these AxiomGT1 genotype calls The SNPolisher Ps Visualization defaults are set to have BB calls as blue upside down triangles AB calls as gold circles AA calls as red triangles The GTC SNP Cluster Graph defaults are set to have BB calls as red triangles AB calls as blue squares AA calls as green circles Note in GTC it is possible to color and shape the data according to other sample attributes which are shown in the legend for the graphs AxiomGT1 genotype NoCalls are made for samples whose Confidence Scores are above the Confidence Score Threshold default 0 15 The Confidence Score is essentially 1 minus the posterior probability of the point belonging to the assigned genotype cluster Confidence Scores range between zero and one and lower confidence scores indicate more confident genotype calls If the Confidence Score rises above the Confidence Score Threshold the genotype call for the sample is converted to a NoCall SNPolisher Ps Visualization defaults are set to have No Calls as gray squares and GTC SNP Cluster Graph defaults are set have No Calls as gray spades The AxiomGT1 cluster variances are used to create ellipses around the cluster means in the SNP cluster plots Ellipses based on priors are dashed and ellipses based on poster
96. t the 96orMore Step2 AxiomGTI1 analysis configuration to perform genotyping for this group of high quality samples gt 96 samples using the generic prior models file and all SNPs on the array Figure 7 7 Panel E f NOTE Select the LessThan96 Step2 AxiomGT1 analysis configuration to perform genotyping for smaller groups of high quality samples 96 samples using SNP specific models file and all SNPs on the array Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 48 Figure 7 7 Overview of steps to perform genotyping on passing samples only using all SNPs and generic priors A B gj Axiom_GW_LAT KNONAQ1006 C05 K NDNAOTOOE C 5 AasomGT1 chp Axiom KP UCSF LAT K NDNAO1007 A01 K NDNAOT007 A lAoxomGTl chp female Ej Sample Attributes 98 KNONAO1007 A02 K NDNAO1007 A 2AxomGTlchp lemale dc Selected Rows to Resuts Group GE Intensity Data 54 KNDNADIOU E06 NDNADI00G E amp AbionGTl cho female Remove Selected Rows from Results Group Py AI 109 K NONAO1007 A12 K NONAQIQU A 12 AxionfTI cho female ace es See ce D In Bounds 112 K NONAQ1007_B04_K NONADIO07_6_4AsscmGT1 chp female ly Out of Bounds 34 K NONAOTO06_C10_K NONAOTO06_C_10AxsomGT1 cho female S E Genotype Results 107 KNONAQ1007 A11 KNONAQIOU7 A 11 AxonGT1 cho female Export Genotype Results 20110923 161145 CHP stepl 4 KNONADIONS H10 K NDNAOT0
97. tep Step 8C 2 eee 35 SNP Call Rate CR o cot aa desc metere manh eu EE Ru EET ee a oe ee 35 Fisher s Linear Discriminant FLD liliis 36 Heterozygous Strength Offset HetSO liliis 37 Homozygote Ratio Offset HomRO lssselesee eae 38 Additional SNP Metrics that may be Used for SNP Filtering llis 39 Hardy Weinberg p val e lt 2 venir ec tae ate Pa be we ds aed bes Rode Aaa bue dde 39 Mendelian Trio Error 2 eee s 39 Genotyping Call Reproducibility sse 39 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT SOT ARC 2o putres dado x RH CR DUI SR ee ara dc ea ae GRA 41 Execute Steps 1 7 with GTC Software 0 0 e 41 GTC Setup REOR TRU ee er aT nr 41 Step 1 Group Samples into Batches in GTC 1 eee 41 Steps 2 and 3 Generate DQC Values and QC the Samples Based on DQC in GTC 41 Steps 4 5 and 6 Generate QC Sample Call Rates QC the Samples and QC the Plates 44 Step 7 Genotype all Passing Samples and Plates from the Same Batch Using All SNPs 47 Execute Step 8 with APT Version 1 16 1 or Higher liisiilsslle sess 49 Locate GTC Step2_AxiomGT1 Step 7 Output Files 0 2 00 0 000000 0 00 ee eee 49 Step 8A Run Ps Metrics 0 0 eee eee 51 Example Ps Metrics Script 2 0 0 teens 51 Step 8B Run Ps Classification 20 0 eens 51 Example Ps Classification Script isses 51 Visualize SNPs and Change Calls if Desired through GTC Plotted C
98. the Step2_AxiomGTI1 output files as input for Step 8 enclosed in blue box in Figure 7 8 The required files are AxiomGTI calls txt and AxiomGTI snp posteriors txt Note that AxiomGTI1 confidences txt and AxiomGTlI summary txt are used with some SNPolisher functions Figure 7 8 The Step2_AxiomGT1 Step 7 output files File Edit View Tools Help Organize v fx Open v Burn New folder Fr Favorites a BE Desktop amp 960r mGT1 GenotypeConsoleAnalysis log Recent Places 9013050 HD oc oc mod d Chromosome Analysis Suite 1 2 1 20130507 141145 CHP AxomGTI1 analysis job 5 dat 20130507 141145 CHP SnpSummuary analysis job AxiomGTI calls tt b TI Libraries AxiomGT1 confidences bit Documents R fooomGTI1 Genotyping Console log 2 Music AxiomGT report txt im Pictures AciomGT1 snp summary db B video AxdomGT1 snp posteriors tt L AxiomGTI summary tit iW Computer NA11839 20130502 A05 AxiomGTI chp The file in the red circle is the Step 2 GTC analysis log and the files in the blue circle are the AxiomGT1 output files To locate the exported text files to use as input for Ps Metric right click the dataset and choose Show File Locations option see Figure 7 9 A use the file location data to identify the genotype result folder see Figure 7 9 B then navigate to the folder and confirm that all Step 2 output files are present see Figure 7 9 C Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT
99. tion and spurious allelic association Lancet 2003 Feb 15 361 9357 598 604 Clayton DG Walker NM Smyth DJ Pask R Cooper JD Maier LM Smink LJ Lam AC Ovington NR Stevens HE Nutland S Howson JM Faham M Moorhead M Jones HB Falkowski M Hardenbol P Willis TD Todd JA Population structure differential bias and genomic control in a large scale case control association study Nat Genet 2005 Nov 37 11 1243 6 de Bakker PI Ferreira MA Jia X Neale BM Raychaudhuri S Voight BF Practical aspects of imputation driven meta analysis of genome wide association studies Hum Mol Genet 2008 Oct 15 17 R2 R122 8 Didion JP Yang H Sheppard K Fu CP McMillan L de Villena FP Churchill GA Discovery of novel variants in genotyping arrays improves genotype retention and reduces ascertainment bias BMC Genomics 2012 Jan 19 13 34 Laurie CC Doheny KF Mirel DB Pugh EW Bierut LJ Bhangale T Boehm F Caporaso NE Cornelis MC Edenberg HJ Gabriel SB Harris EL Hu FB Jacobs KB Kraft P Landi MT Lumley T Manolio TA McHugh C Painter I Paschall J Rice JP Rice KM Zheng X Weir BS GENEVA Investigators Quality control and quality assurance in genotypic data for genome wide association studies Genet Epidemiol 2010 Sep 34 6 591 602 Manolio TA Collins FS The HapMap and genome wide association studies in diagnosis and therapy Annu Rev Med 2009 60 443 56 Pluzhnikov A Below JE Tikhomirov A Konkashbaev A Nicolae D Cox NJ 2008 D
100. trics posterior file input AxiomGT l snp posteriors txt call file input AxiomGT1 calls txt metrics file metrics txt The following parameters are required to run Ps Metrics posterior file The file output by GTC APT which contains posterior summaries of clusters for SNPs usually named AxiomGT1 snp posteriors txt produced by Step 7 genotyping metrics file The name of the output file The user must supply a file name in order to save the output as a separate file which is read in to other SNPolisher functions such as Ps Classification In addition one of the following two parameters 1s required call file The file output by GTC APT which contains genotype calls for SNPs usually named AxiomGT calls txt produced by Step 7 genotyping The call file may use either the labels AA AB BB and NoCall or the numeric labels 0 AA 1 AB 2 BB and 1 NoCall chp files A text file containing a list of CHP files from which to take calls one file path per line CHP files are produced by GTC f NOTE Either the call file or the chp files parameter must be provided in order to run Ps Metrics Computational performance is much better using the call txt file rather than the chp file So if it is an option we recommend using the calls txt file Step 8B Run Ps Classification Example Ps Classification Script To run Ps Classification on the metrics file metrics
101. txt for human samples which will generate the output files in the folder named output in the working directory The Ps Classification program is available in APT versions 1 16 1 and higher Ps Classification species type Human metrics file metrics txt output dir output V ps2snp file lt ANALYSIS FILES DIR gt lt axiom_array gt r lt gt ps2snp_map ps lt ANALYSIS_FILES_DIR gt is full path to the analysis lib file directory Table 1 1 on page 7 The following 8 files will be written to output directory PolyHighResolution txt the table with probe set performance and classification results and seven category files CallRateBelowThreshold txt a Hemizygous txt MonoHighResolution txt a NoMinorHom txt a OffTargetVariant txt Other txt a PolyHighResolution txt The output is discussed in Step SB Classify SNPs Using QC Metrics on page 19 Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 52 The main parameters accepted by Ps_Classification are metrics file The name of the file containing output generated by Ps_Metrics ps2snp file The name of a file containing the mapping of probe set IDs to SNP IDs The column names should be probeset_id and snpid The user must supply this file when classifying SNPs that have more than one probe set This file lt axiom_array gt r lt gt ps2snp_map ps should be provided with the Analysis Library F
102. type calls obtained from the control sample can be compared to the expected genotype calls generating a concordance measurement A low concordance score may indicate that there were either plate processing and or analysis issues Before beginning the laboratory work of processing the human samples investigators should examine the ethnic backgrounds and pedigrees of the proposed samples to ensure there is no population substructure present that could confound the analysis of data from cases and controls e g all of the controls are CEU while the cases are YRI For non human samples the same principles apply and samples should be randomized with regards to breeds species and subpopulations for genome under study In addition researchers should ensure their experiments are sufficiently powered to answer the question of interest Again it is best to examine all of these questions prior to the initiation of the project For a non ideal study design for which cases and controls are not randomized the SNPolisher package provides the BalleleFreq_Test function to identify and remove SNPs with inconsistent genotypes due to shifts in intensity in probe sets across samples that were processed in the separate case and control batches See the Ballele_Freq_Test in the SNPolisher User Guide Pluzhnikov A Below JE Tikhomirov A Konkashbaev A Nicolae D Cox NJ 2008 Differential bias in genotype calls between plates due to the effect of a small number of
103. type retention and reduces ascertainment bias BMC Genomics 2012 Jan 19 13 34 Chapter 3 Best Practices Genotyping Analysis Workflow 25 SNP classification should classify multi cluster SNPs as Other CallRateBelowThreshold or OTV In some cases these SNPs have complex patterns that escape the standard SNP QC filters for these classes If visual examination identifies that multi cluster SNPs are being included in any of the default recommended classes Table 3 3 Supplemental filters can be applied see the SNPolisher User Guide Section on Ps_Classification_Supplemental note that 64 bit Perl should be installed when using Ps_Classification_Supplemental SNPs in the OTV class can be correctly re labeled with four genotype states including OTV see Adjust Genotype Calls for OTV SNPs on page 28 The cluster graphs of the multi cluster SNPs can be examined for possible causes of extra clusters by coloring samples according to different batch variables and or known sample subpopulation structure different breeds lines varieties subspecies etc The by sample coloring option is available in both GTC and SNPolisher software If samples in outlier intensity clusters can be colored based on a common variable for example a common Plate ID or a common sub species the potential root cause may be identified The user may want to repeat Best Practices Step 7 genotyping excluding the samples that form outlier intensity clusters Figure 3 6 Ex
104. ualization function can be used to produce SNP cluster plots Plots include the posterior information default and prior information optional Reference genotypes can also be included The cluster plots can help quality check SNPs and diagnose underlying genotyping problems All plots are output as PDFs The user can adjust the colors used in the plots and highlight select samples For more details on the various options with Ps Visualization see the SNPolisher User Guide Ps Visualization takes six required arguments and eight optional arguments The six required arguments are the name of the ps file with the SNPs for plotting the name and location of the output PDF file the name and location of the summary file the name and location of the calls file the name and location of the confidences file and the name and location of the posteriors file The list of SNPs pidFile can either be the list of SNPs output by Ps Classification for one category files or a list of SNPs selected by the user The first line of the ps file should always be probeset id The user should also give Ps Visualization the name of a temporary directory which will be used for out putting intermediate files temp dir If no directory is given the default used is Temp The accompanying logical operator keep temp dir indicates whether the temporary directory should be kept or deleted at the end of Ps Visualization default is FALSE The user may wish to keep this
105. ud Ve pd eR RE Ire 6 sec ELTE 6 Analysis SOLE 3 iie duction ota o Sages ease the E E hat Angee ene eA 6 ITFOGUGTIOM ERIS 9 Backgro Na Cr 10 Axiom Array Terminology n s asaan aana 10 WEITER TE T eee ete onto om Apia a Seapine ned 10 What is a SNP Cluster Plot for AxiomGT1 Genotypes 2 0 00 00 ee 10 Best Practices Genotyping Analysis Workflow 0 000 eee eee 13 Design the Study to Avoid Experimental Artifacts 2 0 0 0 eee 13 Execute the Required Steps of the Workflow llle 14 Step 1 Group Sample Plates into Batches 0 0 ce eee 14 Step 2 Generate Sample DQC Values 2 0 ee eee 15 Step 3 QC the Samples Based on DQC ee eee 16 Step 4 Generate Sample QC Call Rate Using Step1 AxiomGT1 000 16 Step 5 QC the Samples Based on QC Call Rate llli eee 16 Step 6 OC the Plates as capts Reed hx Ercnnfeemaodnhkntehs wtgnz fuos dea 3 e Pi runi 16 Step 7 Genotype Passing Samples and Plates Over Step2 AxiomGT1 SNPs 18 Step 8 Execute SNP OC bbirr ka eee te dede dd eo ded dete ted ode ed ede 19 Step 8A Create SNP QC Metrics 2 0 a 19 Step 8B Classify SNPs Using QC Metrics lllsslee I 19 Step 8C Create a Recommended SNP List llle 22 Evaluate SNP Cluster Plots 4 cucmd2obiten ob bdda RIDE EIIIISQRIQRbenr dde 23 Well clustered vs Mis clustered SNP Cluster Plot Patterns 0 2 0 0 0 0 0000 000 24 Multi cluster SNP Cluster Plot Patterns n anaana eens
106. up of samples by clicking on the plot and circling the samples with the mouse before releasing the mouse button see Figure 7 15 The lasso function automatically draws a straight line to the starting point of the shape if the mouse button is released before the shape is closed The samples in the group and the rows in the Sample Table are selected when the button is released If the Sample Table is sorted by the Checked column the selected samples will be moved to the top of the table Figure 7 15 Selecting Multiple Samples vu T a Manually Change a Sample s Call To manually change a sample s call click the sample to select it then right click it The Change Call menu appears Select the new call Figure 7 16 Figure 7 16 Manually Changing a Genotype Call Mj Copy image to clipboard la Save image to file Create Custom Intensity Data Group From Checked Samples Create Custom Intensity Data Group Excluding Checked Samples E Scale Change Call gt oo mA 1 s 4 t JU t t s 4 vaea v e lt e Log Ratio Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software 58 Lasso Function The lasso function can be used in a number of different cases including cluster splits In Figure 7 17 the top half shows a cluster split and the bottom image shows the graph after setting the samples correctly to AB Figure 7 17
107. ver contamination will cause QC call rates to decrease Figure 5 1 B shows the effect of deliberately mixing 4 samples enclosed in box Figure 5 1 C includes one plate green points where some samples were accidentally contaminated during pipetting In both plots the contaminated and deliberately mixed samples fall obviously below the curve formed by the uncontaminated samples If the analysis of the DQC and QC call rate correlation pattern of a plate reveals a significant number of samples with high DQC values and low sample QC call rates it may be an indication of sample contamination associated with these samples If the source of sample contamination is understood it s possible to proceed with the study after eliminating just those samples that obviously fall into the contamination zone Note that contamination will produce the pattern in Figure 5 1 C but it has also been observed that large image artifacts on the array surface can produce this pattern as well 1 Laurie CC Doheny KF Mirel DB Pugh EW Bierut LJ Bhangale T Boehm F Caporaso NE Cornelis MC Edenberg HJ Gabriel SB Harris EL Hu FB Jacobs KB Kraft P Landi MT Lumley T Manolio TA McHugh C Painter I Paschall J Rice JP Rice KM Zheng X Weir BS GENEVA Investigators Quality control and quality assurance in genotypic data for genome wide associ ation studies Genet Epidemiol 2010 Sep 34 6 591 602 Chapter 5 Additional Sample and PlateQC 31 Figure 5 1 DC
108. y detect SNPs with error rates significantly higher than the main body of the SNPs the overall error rate is still low in absolute value As discussed in Laurie et al 2010 approximately 30 duplicated pairs of samples are needed to generate enough precision for this type of analysis Discordance rates can also be computed from the 60 samples divided into replicate sets of greater than two In this case a slightly more complicated algorithm is required For each replicated sample set the approach is to first compute a consensus genotype for the sample at the Laurie CC Doheny KF Mirel DB Pugh EW Bierut LJ Bhangale T Boehm F Caporaso NE Cornelis MC Edenberg HJ Gabriel SB Harris EL Hu FB Jacobs KB Kraft P Landi MT Lumley T Manolio TA McHugh C Painter I Paschall J Rice JP Rice KM Zheng X Weir BS GENEVA Investigators Quality control and quality assurance in genotypic data for genome wide associ ation studies Genet Epidemiol 2010 Sep 34 6 591 602 Chapter 6 SNP QC Metrics 40 SNP The number of discordant calls for the sample set equals the number of samples in the set whose genotype does not agree with the consensus genotype The total number of discordant calls for the SNP equals the sum of discordant calls over the sample sets DNA sample quality may vary considerably and these differences in sample quality may influence the genotyping call error rates among samples Therefore the replicated sample sets should be
109. y large homozygous cluster in blue BB divided into several sub clusters The heterozygous AB cluster sits very far below the BB cluster and has a negative HetSO value 0 35 Figure 6 3 D shows a larger homozygous cluster in blue BB and a large cluster that has been split between heterozygous AB calls yellow and homozygous AA calls red This cluster split has caused the true heterozygous cluster to be called as the homozygous cluster This produces a HetSO value of 0 18 The low HetSO values of SNP clusters in Figure 6 3 C and Figure 6 3 D help flag these cases as problematic SNPs Figure 6 3 Examples of SNPs with different HetSO values AB AA 4 BB 2 a e d o o 4 s wo w o o e A T T T 0 2 C HetSO 0 35 D HetSO 0 18 Chapter 6 SNP QC Metrics 38 Homozygote Ratio Offset HomRO If both hom clusters are Min Xaa abs Xbb on correct side of zero If both hom clusters are HomRO Xbb to the right of zero If both hom clusters are Xaa to the left of zero Where Xaa is the center of the AA cluster on the X axis Size Dimension and Xbb is the center of BB cluster on the X axis Size Homozygote Ratio Offset HomRO is the distance to zero in the X dimension from the center of the populated homozygous cluster that is closest to zero If there is only one homozygous cluster HomRO is the distance from that cluster center to zero in the X dimension The heterozygous clu
110. yping Solution produces calls for both SNPs and indels insertions deletions For simplicity in this document the term SNPs will refer to both SNPs and indels Additional chapters in the document include a Chapter 2 Background provides information that is needed for understanding the rest of the document Chapter 3 Best Practices Genotyping Analysis Workflow discusses the required eight steps for producing high quality and appropriate genotypes for downstream statistical analysis as well as guidance on interpreting SNP cluster plots Instructions for executing the steps and visualizing SNP cluster plots are provided in Chapters 7 and 8 Chapter 4 Additional Genotyping Methods discusses methods for changing genotype calls and advanced methods for genotyping more than three genotype clusters Chapter 5 Additional Sample and Plate QC discusses QC considerations for samples and plates that are in addition to those in the required Best Practices steps Chapter 3 a Chapter 6 SNP OC Metrics describes metrics that are used in the Best Practices workflow Chapter 3 for SNP classification as well as additional metrics used in the field for SNP QC a Chapter 7 Instructions for Executing Best Practices Steps with GTC and APT Software provides instructions for executing the seven Best Practices Steps with GTC combined with usage of APT version 1 16 1 or higher for step 8 Instructions for visualizing SNP cluster plots and changing genotyp
111. yping using all SNPs on the array and only high quality samples Using the intensity data group that includes only the CEL files with sample call rates gt 97 after the QC genotyping perform genotyping using all the SNPs on the array Figure 7 7 summarizes the steps to perform genotyping Step 7A Double click the genotyping result batch in the data tree Figure 7 7 Panel A to open the CHP summary table Panel B that is generated after performing the QC genotyping on all the CEL files that pass the DQC cutoff In Bounds Step 7B Identify the outlier samples i e those samples with a sample call rate lt 97 by sorting the CHP summary table by sample call rate Select the call rate column and click the sort ascending icon in the tool bar red arrow in Figure 7 7 Panel B In Figure 7 7 there is one highlighted outlier sample that has a sample call rate below the 97 cutoff To exclude outlier samples select all of the samples with a call rate 97 and then right click in the CHP summary table and choose the Create Custom Intensity Data Group Excluding Selected Samples menu option Provide a name for the newly created intensity data group Figure 7 7 Panel C for example 20110923 165605 CEL outlier excluded after QCgenotyping Step 7C Go back to the data tree and right click the newly created 20110923 165605 CEL outlier excluded after QCgenotyping intensity group to perform genotyping Figure 7 7 Panel D Step 7D Selec

Download Pdf Manuals

Related Search

Related Contents

MODEL DSP – DATA STATION PLUS  Conditions générales de vente de Hottinger Baldwin  USB-DA10  EB82 - iCEstick Evaluation Kit User's Guide  Philips Surge protector SPN4087A  kanomax_2212_indoor_..  V7 LTV32D LCD TV  Sony VPLDX146 data projector  Nextech - Trunnion  PIANO DI SICUREZZA E DI COORDINAMENTO  

Copyright © All rights reserved. Failed to retrieve file