The CNVineta user guide
Contents
1. batch case YRI batch control CEU chrom 5 start 150180253 end 150200253 flanking snps 40 cohort wide TRUE Page 8 of 19 The implemented sorting functionality can even be more helpful to identify samples with LRR shift but that were not called as CNVs Sorting is based on the difference between the mean LRR of the requested region and the mean LRR of its flanks The probe sets of the requested region are labeled with dots Trimming of regions helps achieve the best results In summary using CNVineta we have shown that deletions upstream of RGM are significantly more frequent in YRI compared to CEU and that most YRI individuals have homozygous deletions Some deletions which seem to be heterozygous seem to be missed by the segmentation algorithm and should be examined further CNV burden Calculating the CNV burden load of a sample set is increasingly performed throughout the literature In order to perform a CNV burden load analysis a CNV singleton screen 6 and CNV load statistics are included The singleton screen scans for CNVs which are sample specific and do not overlap with other CNVs This screen is either specific for deletions or duplications or can be used for all copy number states deletions and duplications CNV singleton CNV data obj CNV ineta batch case c Y RI batch control c CEU visualize FALSE The result are provided in three data tables which list the number of samp2. by default same as for dox0 This peak definition is only used for visualization of the data In the following example you will see a significant locus which is covered by several markers representing a fewer number of atoms If all significant atoms in this region were visualized at least 10 graphs for the same region would be produced To prevent this atoms have to be identified which belong to the same set of CNVs Accordingly peaks are defined by the rule described above to reduce the number of regional overview plots do log regression CINV data objCNV ineta batch case c YRI batch control c CEU draw TRUE out file log regression The region on chromosome 11 exemplifies the concept of atoms and peaks The chromosomal region marked by dotted lines consists of several atoms Deletions and duplications can be seen at various positions in different samples However CNVineta identifies the entire region between both dotted lines to be associated Page 12 of 19 4926807 CASS YRY33474 wn 2 a E 5 4917 4922 4927 4932 4937 4942 Chr 11 4916807 4944858 tick marks in kb 53 87 cases with 59 CNV s 4926807 Controls CRY 47 n _ __ 2 a E a 5 4917 4922 4927 4932 4937 4942 Chr 11 4916807 4944858 tick marks in kb 30 88 controls with 49 CNV s In the above example CEU samples are set as cases and YRI samples as controls Significant differences between cases and controls were fou
3. An example command for Affymetrix SNP array 6 0 annotation version na29 located at the folder foo bar could look like this objFileConverter addMarkerAnnotation dbCreator objFileConverter annotation S NP foo bar GenomeWideS NP_6 na29 annot csv header TRUE marker column Probe Set LD chrom column Chromosome pos start Physical Position pos end Physical Position ae FETT sep TOF comment char H na strings objFileConverter addMarkerAnnotation dbCreator objFileConverter Page 17 of 19 annotation SNP foo bar GenomeWideS NP_6 cn na29 annot csv header TRUE marker column Probe Set ID chrom column Chromosome pos start Chromosome S tart pos end Chromosome S top FEE sep kanae ie comment char H na strings As can be seen in the example above the trigger header TRUE is set In this case header names can be given for the required columns If beader FALSE column number instead of names have to be set When using column header names you should keep in mind that R converts special characters to dots That is the reason why marker column Probe Set ID was used instead of marker column Probe Set ID In case of doubt column numbers always work As the next step output file names have to be defined by setting the slot ou file of the objFileConverter object This should be the full path to the binary file that should be created or extended For exam
4. Page 5 of 19 present in YRI while the deletion seems to be absent in CEU In addition the deletion is only covered by non polymorphic probes on the Affymetrix SNP 6 0 array blue ticks while no SNP assays are present in this region The viswalizeRegion function also includes triggers that will output the table and sample names into a file However before further analysis is performed the LRR of the deletion in all samples should be investigated While no SNPs cover this CNV the BAF plot is less important for this example but generated as well The deletion polymorphism is only covered by few non polymorphic probes and might possibly be a false positive CNV due to incorrect calls by the CNV calling algorithm The visualizeRegion function has several triggers which allow for a look at the raw data Setting the out file trigger as well provides graphs in separate files visualizeRegion CN V data obj CN V ineta batch case c Y RI batch control c CEU chrom 5 start 150100000 end 150300000 raw data plots TRUE out file IRGM_overview print tables TRUE Look at all files in the working directory starting with IRGM_overview This will include jpeg files for the regional overview plot raw data LRR and BAF plots of every sample that has a predicted CNV in this region and a table which annotates all predicted CNVs in cases and controls for that region Below is an example plot for an individual of the YRI cohort NA1
5. file stores the LRR and BAF values of every marker To assign the raw data correctly the required columns are marker id LRR and BAF The marker annotation file is a list of all markers on the array and is required to have columns storing information on marker id chromosome start and end Sometimes markers are not annotated with start and end coordinates but merely positions In this case when reading the annotation file into R triggers can be set that allow the R function to use the position value for start and end positions If an annotation file is not available the raw data file 2 can be used instead Usually the raw data file also contains information on chromosome and chromosomal position which is sufficient in combination with the marker id However be aware of inconsistency A raw data file may only have a subset of marker ids which might vary between different raw data files Page 16 of 19 Before continuing with an example conversion process we would like to point out the CNVineta website where three conversion examples are available for download including examples for Affymetrix Power Tools CNV calls QuantiSNP and a Java application to convert Illumina Genomestudio output to the CNVineta format Please visit http www ikmb uni kiel de cnvineta and download these examples from the conversion section In general the Affymetrix Power Tools and QuantiSNP output is very similar and these two examples
6. of exactly 80 000 000 000 byte 74 5 GB The 4 bytes of the R specific NA value missing value in hexadecimal are translated to 00 00 00 80 The segment annotation file is a table of predicted CNVs in the sample set For each CNV information on the individual batch copy number state number of marker covering the segment average marker distance in kb chromosome start position end position start marker and end marker is stored Several well established algorithms are available for CNV prediction and CNVineta is not limited to a specific algorithm The provided HapMap example data set has been analyzed with Affymetrix Power Tools NA18582 CHB 1 0 2 3 31477964 31478724 SNP A 4263432 SNP_A 2315022 NA18521 YRI 3 53 25 22 41227214 41280423 CN_913521 CN_ 913541 Page 15 of 19 Header or comment lines are not allowed in this tab delimited file and the column order from left to right is the same as the column order above starting with individual and ending with end marker Please pay close attention to sample and marker identifiers The individual column has to contain the same sample identifiers as used in the sample annotation file 2 The start marker and end marker entries have to be annotated in the marker id column of the SNP array annotation file 1 5 The RefGene file is a standard list of all genes in the human genome as provided by the UCSC genome browser 3 http genome ucsc edu Since annotations of the human g
7. to be expected since the number of deletions in the control cohort is zero Ignore the warning for now as the vzsualizeRegion function also provides a Fisher s exact test result For a first orientation the visualizeRegion function also generates a regional overview plot This plot stacks all individuals with copy number variations in cases and controls and visualizes the extent of the copy number variation relative to genes Black ticks correspond to SNPs blue ticks indicate non polymorphic probes Deletions are marked in red while duplications are in blue By plotting all individuals with CNVs in cases and controls in a given chromosomal region in a single plot CNVineta provides a quick overview of the segment data in the entire sample set that is used for the association analysis 150100000 cases YRI 150300000 3 i 3 1O Oe Rp p Patat PLEMA O A ETA a ht Paat E I Waho 002 5 C5o0rf62_ IRGM ZNF300__ DCN LOC134466___ 150090 150130 150170 150210 150250 150290 Chr 5 150090000 150310000 tick marks in kb 26 87 cases with 26 CNV s 150100000 controls CEU 150300000 EEE TV a a T A DCTN4 C5orf62 IRGM ZNF300 DCTN4 LOC134466 oOo T T T T T T 150090 150130 150170 150210 150250 150290 Chr 5 150090000 150310000 tick marks in kb 0 88 controls with 0 CNV s As can be seen in the regional overview plot predicted deletions upstream of IRGM are only
8. 9173 LRR a4 7 0 95 4 LRR 2 3 4 Blog ni C5 UEZ IRGM ZNFAQo Dcyi4 LOC134466 T T T T T 150100000 150150000 150200000 150250000 150300000 chromosome 5 pos in bp NA19173 BAF 1 ee A gt gt gt aie a5 4 e s x ao o e o oe Ss DCTN4 CSoxf62 IRGM ZNF B00 DCTN4 LOC134466 T T T T T 150100000 150150000 150200000 150250000 150300000 chromosome 5 pos in bp Page 6 of 19 The upper panel shows the LRR the lower panel shows the BAF in individual NA19173 The red line indicates the average LRR value the red block demarcates the deletion predicted by the algorithm As can be seen the deletion predicted by the CN algorithm is accompanied by a shift in the LRR of several neighboring probes therefore suggesting that the deletion is actually real and not an artifact of the CNV calling algorithm When SNP assays cover a deletion you expect either a loss of heterozygosity heterozygous deletion or total loss of the BAF signal homozygous deletion For duplications the BAF should spread wider The raw data plot inspection as described above is a good way to identify false positives But what about false negatives What if the CNV calling algorithms failed to identify CNVs which are present in some samples but not in others By using the heatmap command the samples with CNVs are compared to random samples in the cohort that do not have predicted CNVs in this regio
9. TRUE out file x0 Page 11 of 19 cases CEU 4191727 4201030 l LI I M l 33 o 25 a 19 E 13 7 i ro e a a a a ep mu ob nma T T T T T T 4191 4193 4195 4197 4199 4201 Chr 2 4191220 4202389 tick marks in kb 33 88 cases with 36 CNV s 4191727 controls YRI 4201030 l I I LI Il M l l HoT UH WU HII PI Win ot Wat T T T T T T 4191 4193 4195 4197 4199 4201 Chr 2 4191220 4202389 tick marks in kb 0 87 controls with 0 CNV s The command above generated a large number of image files starting with x0 in your work directory When screening these graphs each tick at the upper x axis of each graph marks an atom which adheres to the specified x0 rule Genome wide common CNV association analysis The genome wide CNV analysis function of CNVineta works on the same atomic matrix as the x0 analysis However the analysis is conceptually different The genome wide CNV analysis invoked by do og regression initiates a genome wide association study on the elements of the atomic matrix Supplementary Figures 5 and 6 When adjacent atoms or atoms within a defined window size trigger peak window show significant p values these atoms ate merged to one peak This reduces the amount of target regions that have to be visually screened but does not minimize the information A peak table is returned by do og regression
10. The CNVineta Tutorial for the manuscript CNVineta A data mining tool for large case control copy number variation data sets Wittig M Helbig I Schreiber S Franke A Institute of Clinical Molecular Biology Christian Albrechts University Schittenhelmstr 12 24105 Kiel Germany gt Department of Neuropediatrics University Medical Center Schleswig Holstein Kiel Campus Arnold Heller Strasse 3 Building 9 24105 Kiel Germany Contents POGUE Os TEES A E E 1 A sample session using the provided HapMap data Set eeccsscssceseeeeeetseesteeeteenees 2 Visualizing a specific region JRGM in YRI and CEU e cee ceeeeseeneeeeeeeeeeeenseeneeens 4 CNV burde inssin e ea aE E TEE ATENE E E eee eR 9 Genome wide identification of rare CNVS csceseeseeseeseceeeceaecneeeseeeneseeeeaecneeneenes 11 Genome wide common CNV association analysiS cccceccceeseceeseeeseeceeceteeeeeeesaees 12 Generating CNVineta input files fa cianscaccssencatecheiescduennnateiatnesvansnieieeaemiaenterenansdanreese 14 CNVineta input file format description cc eececcceceseceeeceeeeeeececeueceeceeeeesseecsseeneenees 14 Converting CNV calling results to CNVineta input file format cece eeeeeeeeeeeeeee 16 An example conversion PFOCESE si o ceccsseicaiesasisaceaadssetansdeaneisdeieateia daca disssvateveDinoumbentenes 17 Introduction CNVineta is a flexible data mining tool for the analysis of copy number variations CNVs in large
11. allowed and the column order from left to right is sample id zero based index and gender NA06985 0 female NA12003 269 male For the gender column only the entries female gt male and unknown are allowed The binary dat file contains the raw data on LRR and BAF for all samples Depending on the size of the sample set the binary file can be several gigabytes in size However despite the size of the binary file CNVineta was designed with quick access to raw data in mind and data from case control cohorts of thousands of individuals can be visualised rapidly from a standard desktop computer For each marker of each sample two numeric values LRR followed by BAF are stored in binary format The data type is a C float stored in little endian format To access the LRR and BAF of a specific marker of a specific sample the zero based indices belonging to sample and marker and the number of markers on the SNP array have to be known An example Array with 10 000 marker We are interested in the SNP with the zero based index 10 The sample of interest has the zero based index 14 Keep in mind that a C float value allocates 4 byte LRR BAF 8 byte gt LRR read 4 bytes at 10 000 14 10 8 and convert to float gt BAF read 4 bytes at 10 000 14 10 8 4 and convert to float Accordingly if the SNP array includes 1 000 000 markers and the store raw data of 10 000 samples then the binary file must have a size
12. are used to demonstrate how typical CNV calling results can be processed to generate CNVineta input files Illumina data is different as it provided with a large data table which contains raw data and CNV states for the entire sample set in a single file To convert this to the CNVineta format a Java application is provided for the conversion process This Java application can be used for all Genomestudio plugins used for CNV analysis An example conversion process The following commands need to be adapted After loading the CNVineta library create an object of class CNVinetaPreprocessor library CN Vineta objFileConverter new CN VinetaPreprocessor Set a parameter which tells the functions if data should be added to an existing CNVineta input files or if new input files should be generated The following command evokes the generation of new input files objFileConverter add data FALSE Now the CNVineta object is ready for loading the annotation data If the add data trigger is set to TRUE the following commands can be skipped Marker annotation can be added step by step For example Affymetrix provides two separate annotation files for SNPs and non polymorphic probe sets The function addMarkerAnnotation internally uses the generic read table command All triggers available for read table are also available for addMarkerAnnotation allowing for customization of this command Refer to addMarkerAnnotation for more details
13. arker segment and avg marker distance less or equal to 4 kb After this preliminary quality control the data is ready for visualization and further analysis Please be aware to apply well established sample filters before you work with CNVineta Le identification of duplicate or related individuals as well as ethnic outliers should be performed prior to subjecting your data to CNVineta Also the CNV prediction tool specific filters such as MAPD and contrastQC filter for Affymetrix Power Tools analyses or mean LRR and mean BAF scatterplots for QuantiSNP analyses should be applied Visualizing a specific region IRGM in YRI and CEU A specific chromosomal region can be visualized using the viswalizeRegion function This function is an essential function within the CNVineta package as other functions such as the initial visualisation step for genome wide association screening use this function For the HapMap example data set we will use a deletion polymorphism upstream of IRGM which has been shown to be associated with Crohn s disease 2 This analysis will produce a table with the summary of CNVs found in a specific chromosomal region and a regional overview plot In this example case status is assigned to YRI upper panel in Vineta plot and control status to CEU lower panel visualizeRegion CNV data obj NV ineta batch case e Y RI batch control c CEU chrom 5 start 150100000 end 150300000 261 samples to proc
14. ation based linkage analysis American Journal of Human Genetics 81 5 Lao O et al 2008 Correlation between genetic and geographic structure in Europe Curr Biol 18 16 1241 8 6 Zhang D et al 2008 Singleton deletions throughout the genome increase risk of bipolar disorder Mol Psychiatry 14 4 376 0 Page 19 of 19
15. case control SNP array data sets The tool is available as an R statistical package CNVineta offers a flexible and fast access to CNVs by allowing for a quick graphical overview in large case control datasets In addition CNVineta provides rapid access to the log of raw data ratios LRR and B allele frequencies BAF of specific or all samples thereby allowing for a fast verification of the underlying raw data CNVineta is also equipped with analysis methods for genome wide screening for associated rare as well as common CNVs Hence CNVineta is a unique data mining tool to rapidly explore CNVs in large case control data sets A sample session with CNVineta The next pages guide the user through a sample session which uses the free available Page 1 of 19 HapMap data set 1 The sample session starts with CNVineta initialisation followed by sample filtering association screening for rare and common CNVs Finally downstream verification of the predicted associations is included as well Download the Affymetrix 6 0 HapMap data set ready for CNVineta from http www ikmb uni kiel de cnvineta Unpack the splitted zip or tar gz archives and remember the path where the unpacked HapMap data is on your computer For example if files are unpacked at foo bar e g Windows E test remember the directory foo bar CNVineta e g Windows E test CNVineta If the 2 7 GB HapMap CNV dataset should not be downloaded you may use the reduced data wh
16. ed in the analysis For more details use the help functionality do log regression within the R package Keep in mind that it is necessaty to clean your data sample set before doing data mining Generating CNVineta input files CNVineta is a post processing and data mining tool and is capable of analyzing data from various different platforms once the data has been converted to the CNVineta input file format Data conversion functionality is implemented in the CNVineta package and described later on in this tutorial We start with a description of the CNVineta input file format Users may choose to either use the implemented R functionality with all the advantages of generic R functions like read table or convert data using custom made scripts executables for best runtime performance For questions or special solutions please refer to http groups google com group cnvineta CNVineta input file format description In order to provide both a quick overview of the overall CNVs in case control data sets as well as access to the LRR and BAF raw data of individual samples CNVineta requires a specific file format An example data set which has already been converted to the CNVineta format is available at the CNVineta website CNVineta requires the following 5 files 1 The SNP array annotation file is needed to assign genomic coordinates to specific markers and to locate the raw data in the binary file 3 The tab delimited file contains in
17. enome are subject to change CNVineta is not limited to a specific build of human genome the HapMap example uses UCSC hg18 The RefGene file is provided separately rather then connecting to online database allowing for CNVineta to function as a stand alone application Converting CNV calling results to CNVineta input file format For Illumina data please follow the instructions provided in Supplementary Figure S3 CNV calling tools usually provide at least two file types which are necessary to generate CNVineta input files These include a segment report which lists all copy number variation that were called in the sample set 1 and a raw data file which includes the raw data LRR and BAF of all markers 2 Sometimes the raw data file is already used as input for the CNV calling tool e g QuantiSNP With all other CNV calling algorithms both file types have to be provided to make the data compatible for CNVineta The array annotation file 3 is usually provided by the array manufacturer If not provided the required information can usually be derived from the raw data files 2 as well These are input files required for CNVineta in detail 1 The copy number segments file stores information about all CNVs which were called by the CNV calling algorithm This file includes information about the chromosomal location of each CNV as well as start and end coordinates The fourth required column is the copy number state 2 The raw data
18. ess 26 samples with CNV 1 cases 26 segments in 26 samples 1 controls 0 segments in 0 samples 1 got 7 transcripts size 01 2 3 4 del samples dup samples CNP samples YRI 87 24 2 NA 0 0 26 0 26 sum 87 24 2 NA 0 0 26 0 26 size 0 1 2 3 4 del samples dup samples CNP samples CEU 88 00 NAO 0 0 0 0 sum 88 00 NAO 0 0 0 0 Deletion Duplication Copy Number Polymorphism Pearson s Chi squared 9 034387e 08 0 9 034387e 08 Fisher s Exact 1 406624e 09 1 1 406624e 09 1 150100000 150300000 Warning message In chisq test chi_table Chi squared approximation may be incorrect In summary this analysis detects a strong association of a deletion assigned to the case cohort YRI In fact 26 out of 87 individuals of this cohort have a predicted deletion while Page 4 of 19 none out of 88 individuals in the CEU group has a deletion Just be aware that in this example the association analysis is performed over the entire region on chromosome 5 start 150100000 end 150300000 and the association analysis does not take into account the location of the deletions or whether there is overlap between CNVs It is a fundamental problem for any CNV association analysis that p values do not provide information on the relative location of the segments We will provide a solution with the later described functions dox0 and do log regression In the above example R produces a warning concerning the Pearson s Chi squared test This is
19. formation on marker name a zero based numeric index for internal usage as well as chromosome and physical start and end position The SNP array annotation file is mandatory for allowing CNVineta to assign raw data locations in the binary file by using the zero based numeric index No header or comment lines are allowed and the column order from left to right is marker id chromosome start position end position and zero based index SNP_A 4261452 2 7430093 7430093 251547 CN 922409 22 49581309 49581334 1880793 Marker names starting with CN will be identified as non polymorphic probe sets marker names starting with SN or rs as SNP probe sets If your probe set annotation does not include start and end coordinates but position use the same entry for start and end The markers must be ordered by chromosome followed by start and end coordinates The order of chromosomes is itrelevant but all markers of the same chromosome must be in one block and ordered by position The zero based index should be applied to the sorted marker table Page 14 of 19 2 3 4 The sample annotation file is a list of all samples included in the analysis The tab delimited file stores information on sample name a zero based numeric index for internal usage and the gender The sample annotation file is mandatory for allowing CNVineta to assign positions in the binary file 3 by the zero based numeric index No header or comment lines are
20. ich is provided with the package Please read the first paragraphs of the sample session until the discussion of the f erSampks command which will explain the handling of the reduced dataset To start your first CNVineta session download the CNVineta package from http www ikmb uni kiel de cnvineta and install the package in your R framework A sample session using the provided HapMap data set First use the generic R command setwd and set the working directory to the folder which contains the HapMap data After this command all further R commands in this sample session ate executable by copy and paste Remember the directory where you have unpacked the HapMap data and execute setwd setwd foo bar CNVineta If you want to go through the tutorial without copying and pasting the commands step by step you can run a script which does the entire CNV screening process To do so please paste the following command source example_cmds R To perform the automatic screening by using the example dataset provided with the package please execute library CNV ineta ranCNV inetaExample my output is here Now you can start with the step by step tutorial After installation the CNVineta package is loaded with library CN Vineta For the following analysis steps a CNVineta object has to be generated If the reduced dataset is used which is provided with the package initialisation is slightly different and the following
21. les number of singletons singletons per sample number of samples with singletons and number of Page 9 of 19 singletons overlapping genes for the three different CNV singleton types 1 singletons sample number singleton number singletons per sample cases 87 309 3 551724 controls 88 205 2 329545 samples with singletons singletons overlapping genes cases 82 122 controls T5 98 1 del singletons sample number singleton number singletons per sample cases 87 291 3 344828 controls 88 164 1 863636 samples with singletons singletons overlapping genes cases TI 102 controls 70 71 1 dup singletons sample number singleton number singletons per sample cases 87 294 34379310 controls 88 139 1 579545 samples with singletons singletons overlapping genes cases 51 238 controls 52 93 If the parameter visualize is set to TRUE regional overview plots are generated automatically We suggest also setting the parameter out file because a large number of singletons can be expected per analysis The CNV load functionality assessing the number of segments per sample is already included in the filtering process where filtering is based on the number of segments per samples To combine different batches to cases and or controls and calculate the case control specific CNV load you can use the CNV ad function CNV load CN Vdata obj CNV ineta batch case c Y RI batch control c CEU The function produces following o
22. mand and set a path with setwd to the location where the output files should be exported to obj CN Vineta initializeCN VinetaExample setwd my output is here SNP array data analyzed for CNVs usually shows a wide variation of copy numbers per sample CNVineta can identify samples with excessive copy number segments per sample Implemented filter criteria are either w for discarding samples with boxplots whiskers as threshold Or q with the quantiles as threshold The following command will generate a boxplot with the number of segments per sample in each batch before and after filtering as well as a summary table including information on the number of samples included in each batch before and after filtering obj CN Vineta filterSamples obj CN Vineta out file filter 1 after filtering segments per sample batch samples original samples filtered delta delta percent 1 CEU 90 88 2 LZ 2 CHB 45 43 2 4 4 3 JPT 45 43 2 4 4 4 YRI 90 87 3 3 83 In this analysis 2 out of 90 CEU samples were discarded Two graphs will appear in the working directory use getwd to determine the actual working directory which show batch wise boxplots before filter01 jpg and after filter02 jpg filtering Page 3 of 19 segments per sample segments per sample 1000 number of segments 3162 number of segments 63 1 100 50 39 8 at least 5 marker segment and avg marker distance less or equal to 4 kb atleast 5 m
23. n visualizeRegion CNV data oli CNV ineta batch case Y RI batch control c CEU chrom 5 start 150100000 end 150300000 add heatmap TRUE Page 7 of 19 PEP EPPEEPE PEELE PEEPEPE PEPE T mr EAEE NSARE EEEIEE cess R EEEE EENEN TEENER EEE AREE PSE E SESU EEE OSEP AETR 33m e ak E 828 3 WS MISS marker 33 BR 3 In this heatmap case samples with predicted deletions are indicated by three symbols control samples with predicted CNV are indicated by one symbol not present in above dataset Changes in the LRR are indicated by color changes red lower LRR As can be seen there is a strong difference between individuals with deletions and individuals without deletions However there are several individuals which appear to have a deletion These deletions however were not called by the CNV calling algorithm indicating that some samples might have CNVs that were not identified by the CNV calling algorithm Therefore checking the entire cohort in a cohort wide heatmap should be the next step A heatmap for the entire sample set helps to identify the number of false negatives A heatmap sorting function is also implemented in order to trim the query to cover only the CNV region in question This generates a heatmap which has better sorted rows The trigger flanking snp defines the number of SNPs that should be added left and right to the requested region in the heatmap getHeatmap CN Vdata obj CNV ineta
24. nd for various regions demonstrating that frequencies of CNVs in this regions varies significantly between various ethnicities To reduce signals resulting from population stratification covariates can be added to the logistic regression model Within the example data set an additional file with eigenvectors for each sample is provided These eigenvectors were generated with plink 4 In the following example the first two of the four calculated eigenvectors were used as covariates do log regression CN Vdata obj CNV ineta batch case c Y RI batch control c CEU draw FALSE cov header TRUE cov idx 1 cov columns dim1 dim2 cov sep Nt covariates CN Vineta HapMap Affy6 0 eigenvectors txt Adding these covariates to the model the corrected p values are close to 1 Accordingly the Page 13 of 19 previously observed differences between CEU and YRI seem to be associated with the eigenvectors which in this case are correlated with the geographic origin of the samples 5 The covariates table can be added in different ways as a separate file as outlined above or as a separate data frame table The parameter cov idx provides the do ogregression function with the column containing the sample ids which can be a numeric index or a character depending on the column header name The same applies for the cov columns parameter which provides information on columns containing the covariate that should be includ
25. ple if a binary file called Test binary located at the folder foo bat should be created use the following command objFileConverter out file foo bar Test binary Please choose the filename careful If pointed to an existing binary file with the trigger add data FALSE all files will be deleted before file conversion starts If data should be added to an existing file set the trigger add data TRUE In addtion objFileConverter out file has to point to the binary file including all other CNVineta input files in the same folder The final step before executing createCNVinetaDb is setting up a table which provides information on the location of segment files 1 and raw data files 2 and includes information on sample ID gender and batch This table should have the following column names raw data file seoment file sample name sample gender and batch Column entries should contain the corresponding information A tab delimited example file called example txt located at foo bar could look like this raw data file segment file sample name sample gender batch data raw 1 dat data CNV 1 txt Samplel female ShipmentA data raw N dat data CNV N txt AnotherName male ShipmentB The following command will load that file to the slot she data table of the objFileConverter object objFileConverter the data table read table foo bar example txt sep t header TRUE Now everything is ready to start the data conversion
26. three commands can be skipped continue after samples2batches obj CN Vineta new CNV ineta This object called objCNVineta in this tutorial will combine information on the input files binary dat file SNP array annotation file sample annotation file predicted CNVs file and RefGene file Navigate to the directory with the sample files as described above setwd or use full pathnames instead of filenames only and initialize the CNVineta object using Page 2 of 19 initializeCN Vineta This might take a minute or two depending on the SNP array and sample number obj CN Vineta initializeCN V ineta objCN Vineta sup anno CN Vineta HapMap Affy6 0_binary dat snp idx tet ind anno CNV ineta HapMap Affy6 0_binary dat sample idx txt binary dat CNV ineta HapMap Affy6 0_binary dat segments orig CN Vineta HapMap Affy6 0_binary dat segments txt ome IF refgene anno refGene txt For the subsequent analysis different samples may have to be assigned to new or other batches Information on batches which will also be used for case control analysis later on has to be provided in a tab delimited file which has two columns with headers batch and sample Type P samples2batches in R for an example question mark followed by a command name opens a help page obj CN Vineta samples2batches obiCN Vineta samples2batches txt For the reduced package data set please start with the following com
27. utput samples median deletion 87 52 88 53 median CNV 72 0 66 5 median duplication 19 14 cases controls Page 10 of 19 deletions per sample duplications per sample CNVs per sample 80 25 80 60 20 70 50 15 i 60 L 40 10 30 i T T T T T T cases controls cases controls case controls Genome wide identification of rare CNVs CNVineta offers the possibility to identify CNVs which are only found in cases but are absent in controls We refer to this analysis as XO analysis x deletions or duplications in cases vs 0 in controls and it is invoked by the dox0 command The basic idea of the genome wide CNV analysis is reducing the genomic positions for analysis to a minimum CNVineta defines so called atoms Supplementary Figure S2 which represent marker positions where a predicted CNV in any sample either starts or ends These atoms are then used for the entire sample set to create a atomic matrix for the genome wide scan The x0 analysis simply looks for atoms which have CNVs in cases but not in controls Supplementary Figure S4 The graph below shows a deletion on chromosome 1 which is only found in YRI but not in CEU A complete table with these regions can be generated using specific triggers for dox0 dox0 CNV data obj CNV ineta batch case c CEU batch control c Y RI draw TRUE min diff 5 max one side 1 cases more affected
28. with the command createCNVinetaDb For a detailed description of the various triggers for that command please type PereateCNVinetaDb in your R command prompt An example command which uses the most important triggers could look like this createCN V inetaD b dbCreator objFileConverter raw marker column Probe S et raw LRR column Log R Ratio raw BAF column B allele Frequency seg chrom column Chromosome Page 18 of 19 seg start column Start bp seg end column FEnd bp seg state column Copy Number raw header TRUE raw sep Nt seg header TRUE seg sep 2 For both input file types i e segment files 1 and raw data files 2 all triggers available for the generic R command read table are also available via createCNVinetaDb This allows customization of the specific input file format For segment files the prefix seg and for the raw data files the prefix raw should be used as indicated by the arguments vaw header and seg header References 1 International HapMap Consortium The International HapMap Project Nature 2003 426 6968 789 96 2 McCarroll S A et al 2008 Deletion polymorphism upstream of IRGM associated with altered IRGM expression and Crohn s disease Nat Genet 40 9 1107 12 3 Karolchik D et al 2009 The UCSC Genome Browser Curr Protoc Bioinformatics Dec Chapter 1 Unit1 4 4 Purcell S et al 2007 PLINK a toolset for whole genome association and popul
Download Pdf Manuals
Related Search