Parallel Optimization Workbench (POW) - User Manual -
Contents
1. appended in a unique numpy array e write_pdb outname save a new PDB file containing all the monomers treated as chains of the same assembly 7 2 2 Search Space The conformational space of rigid assemblies having a circular symmetry is defined by the three rotation angles a 6 y of a single monomer with respect of a center of symmetry aligned along the z axis and a displacement r with respect to it which represents the radius of the assembly in its narrowest point If an ensemble of ligand structures is available obtained for instance from a MD simulation or alternatively NMR or X ray experiments flexibil ity or multiple conformations can be introduced as set of further dimensions in the search space To do so a principal component analysis PCA is ini tially performed on the ensemble The projection value of every trajectory frame along the most relevant eigenvectors also called fluctuations is com puted These are used as a way to index the trajectory frames which we can consider as a protein conformation database This module can also flexibly or rigidly assemble a multimeric complex around a rigid receptor In this case four additional degree of freedom i e the translation of the whole assembly along the z axis and the three rotations 6 0 4 of the receptor around itself In summary the search space dimensions are in order Qa p Vr Z Q 0 Dy eig_1 eig_2 7 2 3 Fitness Function The fitness function sc2. framework is represented in Figure 1 Every box corresponds to a specific class Classes highlighted in blue are common to any optimization problem and can be considered as a black box by the user Classes in the yellow area change depending on the problem being solved We will call module a file containing an implementation for these classes aiming at solving a specific problem In order to use POW a user has to provide two information the module name and a parameterization file The parameterization file contains a set of keywords associated to one or more values Some keywords are standard for any optimization problem whereas others are problem specific The classes DefaultParser for stan dard keywords and Parser for custom ones are in charge of reading the input file Once the parameters are parsed POW loads if needed specific data struc tures required by the user This operation is performed by the class Data Since this class is part of a module depending on how this class is imple mented any data structure can be manipulated Subsequently POW defines the problem s search space Every dimension of the search space is defined by upper and lower boundaries as well as by spe cific boundary conditions Creation of the search space is problem specific Default Parser Default Space input o module PSO result Default Postprocess Figure 1 Schematic of the Optimizer architecture Every box
3. indeed match the order of target 18 measures provided with the target keyword in input file The multimer pa rameter is a Multimer object see 7 2 1 This object provides the following functions for measurement of the structure e multimer get_width returns the assembly width e multimer get_height returns the assembly height e multimer atomselect unit chain resid name returns a numpy 2D array containing all the coordinates of atoms matching the selection e multimer distance a b returns the minimal euclidean distance within two ensembles of points a and b numpy 2D arrays returned for instance by the atomselect keyword 7 5 Parameterization Examples The minimal set of keywords for a POW parameterization file for protein assembly are as follows monomer input pdb constraint constraint py degree 5 radius 10 target 10 20 This will rigidly assemble 5 monomers from file input pdb so that the circu lar radius is 10 constraint py file will be used as constraint This file will compute two measures that should be compared with the target measures 10 and 20 A complete example showing how to perform a rigid assembly is as follows steps 150 particles 50 repeat 3 boundaryMin 000 8 boundaryMax 360 180 360 12 assembly_style rigid 19 monomer protein pdb constraint constraint py degree 7 target 85 150 filter _threshold 0 cluster_threshold 5 In this example a calculation protocol with 150 iteration
4. of specific atom selections If instead of a simple PDB file an ensemble of structures is provided POW will create a PDB called protein pdb which will be parsed and subsequently used as an index The ensemble of structures will be saved as a set of alternate coordinates In detail the following methods are implemented e import_pdb pdb parse a PDB file e coords get_xyz get cartesian coordinates of every atom Returns a numpy Nzx3 array where N is the number of atoms e set_xyz coords set cartesian coordinates of every atom coords must be a Nx3 array where N is the number of atoms 11 rotation x y z rotate the protein according to angles around the x y and z axis r rgyr compute gyration radius r c center compute geometric center c coords atomselect chain resid atom get cartesian coordinates of a subgroup of atoms selected by their chain name chain residue id resid and atom name atom Returns a numpy M23 array where M is the number of slected atoms Chain resid and atom can be also a wildcard symbol selecting all atoms write_pdb outname save a new PDB file Note that in order to speedup the calculation and simplify data storage Protein stores a PDB as a numerical numpy array Every chain name atom name and residue name are converted into a numerical equivalent using a dictionnary The Multimer class is responsible of assemblying multimers on the base of an initially given Protein object At th
5. Default value 0 Description An ensemble of solution is found but just some of these will be good This variable sets a threshold on the solutions fitness function output lt text file gt Acceptable values UNIX filename Description The text file will be used to store results restart_ freq lt restrart writing frequency gt Acceptable values int Default value round steps 10 Description POW can automatically generate restart files saving the swarm state These can be used to restart the optimization process after a crash Setting this variable to 1 will disable restart writing 6 save restart lt restart saving file name gt Acceptable values UNIX filename Default value swarm restart Description Name of the restart file POW will automatically save at a frequency given by restat_freq The restart contains informa tion about timestep repetition particles positions velocities as well as position and value of their respective current best solution During ex ecution both the most recent restart and an older copy of it are stored default name swarm restart old load _ restart lt restart loading file name gt Acceptable values UNIX filename Description By providing a restart file the optimization process will restart from the last saved timestep and repetition When restarting an optimization the original input file should not be changed the addition of load_ restart statement is sufficient The pre
6. Parallel Optimization Workbench POW User Manual Laboratory For Biomolecular Modeling Institute of Bioengineering School of Life Sciences Ecole Polytechnique F d rale de Lausanne EPFL Lausanne Switzerland Contents Requirements Architecture Provided Files A Launching Standard Keywords Function Module DockSymmCircle module a br AAA 7 2 Implementation Details 2 444 4246 dass 7 2 1 Data Steele Le me 24e sod oe we ea ae ee ee he eee ee ae ee eee ee eee 124 CRG e s c a oe OR A aa et AROS s siae a wi ARE AN 7 4 Constraint Filel 7 5 Parameterization Examples 2244 4 Lu dse6 nS Hie we 8 Creation of a new POW Module 10 10 11 11 13 13 14 15 18 19 21 1 Requirements POW requires the following python gt 2 5 packages to be installed e numpy e mpi4py The execution of parallel calculation will also require the installation of Open MPI Additonal packages may be required by specific POW modules e scipy used by the modules DockDimer and DockSymmCircle e MDAnalysis used by the modules DockDimer and DockSymmCircle e wxpython required when running the GUI of the Function module 2 Architecture POW is a framework allowing the resolution of virtually any optimization problem via the addition of a specific module This object oriented code is developed in Python and supports parellel computation by exploiting MPI libraries The architecture of our
7. aMax lt maz inertia of particles gt Acceptable values float 0 1 Default value 0 9 Description It is the maximum inertia of particles Between steps of the PSO the inertia is decreased until inertiaMin e inertiaMin lt min inertia of particles gt Acceptable values float 0 1 Default value 0 4 Description It is the minimum inertia of particles Between steps of the PSO the inertia is decreased until inertiaMin cp lt influence of local best solution gt Acceptable values float Default value 1 2 Description It is the influence on a particle of the best solution found by that particle cn lt influence of global best solution gt Acceptable values float Default value 1 4 Description It is the influence on a particle of the best position found by neighbors of that particle kar_ threshold lt threshold for kar execution gt Acceptable values float gt 0 Default value 0 01 Description When a particle is being slower than this threshold the kick and reseed procedure KaR will be triggered The particle will receive a random kick that will reaccelerate it If moreover the particle s current fitness is smaller than filter_ threshold it will be also reseeded in a random location This avoids early convergence and forces the swarm to explore further the search space Notice that setting kar_ threshold to 0 disables KaR filter_ threshold lt fitness value to accept gt Acceptable values float
8. are not sensitive to variations of this value The rough energy function in equation 2 only avoids clashed of subunits and at the current stage is not sufficiently precise to allow a blind docking i e a docking where no geometric restraints are provided However work in the development of more accurate energy functions to be included in the fitness function is currently ongoing We expect this will enhance the capabilities for the broad problem of protein protein recognition 7 2 4 Clustering All fitness evaluations obtained during PSO are collected and solutions hav ing a fitness lower than a predefined threshold are retained In most applica tions the filtering criteria is set to 0 Such a value indicates that most likely the system s energy is negative and geometric restraints are well respected Since several solutions usually represent similar conformations clustering is performed Two ad hoc clustering approaches able to determine automati cally the number of required clusters are available the first groups solutions 14 being close enough in the search space preimplemented in Default py whereas the second clusters solutions generating assemblies having a small RMSD within themselves Cluster representatives are selected cluster cen ters ranked according to their fitness and their corresponding assemblies returned as an ensemble of PDB files 7 3 Keywords Additionally to default POW keywords the following keywords ar
9. e assembly is displaced along the z axis with respect of the receptor Boundary conditions are defined by a lower and higher boundary These are computed around the size of the receptor z_padding adds an additional dislacement to the computed boundaries Should be defined only if boundaryMinReceptor and boundaryMaxReceptor are undefined and if receptor is given boundaryMinReceptor lt min boundary for receptor dimensions gt Acceptable values list of lower boundary for each dimension sepa rated by spaces Default value min_receptor z_ pad 0 0 360 2 degree Description It is the minimum boundary for each dimension of the space The first three values correspond to the rotations of the monomer on x y and z axis respectively The last one is the value specified by the radius keyword In case you did not use the radius keyword you MUST specify a minimum radius here boundaryMaxReceptor lt max boundary for receptor dimensions gt Acceptable values list of upper boundary for each dimension sepa rated by spaces 17 Default value max_receptor z_pad 0 0 360 2 degree Description It is the maximum boundary for each dimension of the space The first three values correspond to the rotations of the monomer on x y and z axis respectively The last one is the value specified by the radius keyword In case you did not use the radius keyword you MUST specify a maximum radius here e cluster_threshold lt clustering distance within
10. e creation of a new module i e a spe cific implementation of the Parser Data Space Fitness and Postprocess classes is trivial even for a user unaware of its internal architecture The following modules are already available e DockDimer dock two proteins into an heterodimer e DockSymmCircle rigid flexible assembly of n monomers according to a circular symmetry possibly in presence of a receptor e Function generic function optimization In the next sections these modules will be described 3 Provided Files The compressed folder POW tar gz containing all the needed files is down loadable at lbm epfl ch resources This file unpacks in a folder called POW which can be placed anywhere in your computer The folder contains the following files e Assembly py data structure for heterodimers assembly e Default py classes common to any POW implementation e DockDimer py dock two proteins e DockSymmCircle py rigid flexible assembly of n monomers according to a circular symmetry possibly around a given receptor e Function py generic function optimization e flexibility py functions for Principal Components Analysis e parse py performs just the postprocessing without running PSO This is useful when POW has been already run and just alternate postprocessing options on the produced results have to be tried Usage goes as follow parse py module input_file logfile e POW py main executable e Protein py PDB par
11. e defined e radius lt fixed radius of the multimer gt Acceptable values float Description use this keyword if you know precisely the multimer internal radius in its narrowest point If the precise radius is not known the user should define reasonable boundaries for the pore radius value via the boundaryMin and boundaryMax keywords e degree lt number of monomer gt Acceptable values positive integer Description It is the number of monomer that compose the multimer e target lt list of measures gt Acceptable values list of float separated by spaces Description The list of target measure will be used by the system to compute the fitness This list MUST have the same schema as the list computed from the constraint file e constraint lt constraint file gt Acceptable values UNIX filename Description The system generates a multimer corresponding to the particle position in the space and passes it to the constraint file See the section for details about the structure of this file The list of measure you return will be compared to the list of target s measure and output a fitness value that will be written to the output file The list of target measure MUST have the same order as the list of measure computed in the constraint file e style lt type of assembly gt Acceptable values flexible rigid Default value rigid Description Define the type of assembly to perform If rigid is cho sen the monomer keyword mu
12. e moment this class can just produce multimers according to a circular symmetry In detail the following methods are implemented create_multimer degree radius pos creates a circular multi mer composed of degree monomers having an internal radius equal to radius and having every monomer rotated according to pos x y z Notice that pos should be a numpy array This is the first method to call after initialization A list of degree length of numpy arrays con taining a copy or Protein coordinates is created Subsequently every element in the list is individually manipulated to create a multimeric arrangement multimer_to_origin move the whole complex to the origin z_to_origin move the complex to place its center of geometry at a 0 cords atomselect unit chain resid atom get cartesian coordi nates of a subgroup of atoms selected by their unit id unit numbering of individual monomers counted clockwise chain name chain residue id resid and atom name atom 12 e w get_width get multimer width w e h get_height get multimer height h e d distance select1 select2 compute the minimal euclidean dis tance between two sets of points select1 and select2 e coords get_multimer_uxyz extract coordinates of all atoms in the multimer in a list of length Nx3 numpy array everz element of the list being an arraz of monomer coordinates e coords get_multimer_xyz extract coordinates of all atoms in the multimer
13. minpos maxpos maxvel type keyword 1 180 180 periadic Boundaries Editor 2 180 180 100 periodic test FRE im min 180 max 180 4 180 180 periodic a lt lt max vel boundary typ periodi m f keyword test Add Edit Clone Remove i moras res OK Cancel PSO BEHAVIOR _ Neighbor style geographic Neighbors nb 3 inertia min 0 3 inertia max 0 9 personal weight 1 2 global weight 1 4 O use repulsion field LSE SH Launching module Steps 200 Particles 40 Repetitions 1 Pick how many CPU to use 8 em Can J Fu ons Figure 2 POW graphical interface for Function module allowing the user to save edit validate and launch POW input files 7 DockSymmCircle module 7 1 Overview With this POW module we aim at finding quickly a reasonable prediction for a multimeric structure arrangement on the basis of structural information about its subunits and experimental measures acting as search restraints In a first step an ensemble of monomer conformations is generated typically from molecular dynamics simulations or structural biology experiments this will be treated as a conformational database see Material and Methods The advantage of such an approach is that assembly prediction is performed 10 using physically plausible structures Upon definition of a list of geometric restraints and a specific symmetry a Particle Swarm Optimi
14. ng to the desired number of eigenvectors used for projection Requires style keyword set to flexible align lt define whether to align the given trajectory gt Acceptable values yes no Default value yes Description If set to yes the provided trajectory will be aligned on the protein Taken into account only if style keyword is set to flexible ratio lt energy represented by eigenvectors gt Acceptable values float 0 1 Default value 0 8 Description After having performed PCA POW selects a number of representative eigenvector These will represent at least a certain 16 percentage of the trajectory s energy Taken into account only if style keyword is set to flexible detectClash lt clash detection switch gt Acceptable values on off Default value on Description define whether a 9 6 Lennard Jones function should be computed to assess the system s energy mixingWeight lt weight energetic vs geometric contributions gt Acceptable values float 0 1 Default value 0 2 Description fitness function is computed via the equation f cx energy 1 cx distance where c is the value of mixingWeight receptor lt clustering distance within solutions gt Acceptable values UNIX filename Description PDB file containing a receptor around which the assem bly will be built z_ padding lt assembly vertical displacement gt Acceptable values float gt 0 Default value 5 Description the whol
15. oring the quality of an assembly depends on two fac tors geometry and energy As geometric contribution specific measures of the current multimer m are compared to target values t being experimentally 13 known The aim is to minimize the difference within the obtained and desired values Target measures can be as diverse as width or height obtained from cryo EM maps to atomic distances obtained with FRET or cross linking ex periments Let c m an ensemble of measures performed on a multimer The geometric score G m of a multimer is determined by the euclidean distance within obtained and target measures Glm y E cm Fem a In order to avoid steric clashes during assembly a coarse energy potential is also taken into account This minimalistic contribution is constituted by a 9 6 Lennard Jones type of potential describing all the Ce and Cg atoms of two neighboring monomers extracted from the assembly sm e 2 9 o where r are all the distances within couples of atoms being at a distance smaller than 12 A and e 1 and o 4 7 The values of these constants correspond to the coarse grained parameterization for C atoms in the Mar tini force field The final fitness function f mixes geometric and energetic contributions by means of the following weighted sum f m c E m 1 c G m 3 where c is a real value within 0 and 1 In our tests we set c 0 2 After preliminary tests we found however that results
16. pace is considered e neighborSize lt number of neighbor gt Acceptable values positive integer Default value 1 Description NeighborSize defines the amount of neighbors taken into account by every particle e boundaryMin lt min boundary for each dimension gt Acceptable values list of lower boundary for each dimension sepa rated by spaces Default value module dependent Description It is the minimum boundary for each dimension of the space The first three values correspond to the rotations of the monomer on x y and z axis respectively The last one is the value specified by the radius keyword In case you did not use the radius keyword you MUST specify a minimum radius here e boundaryMax lt max boundary for each dimension gt Acceptable values list of upper boundary for each dimension sepa rated by spaces Default value module dependent Description It is the maximum boundary for each dimension of the space The first three values correspond to the rotations of the monomer on x y and z axis respectively The last one is the value specified by the radius keyword In case you did not use the radius keyword you MUST specify a maximum radius here e boundaryType lt type of the boundary gt Acceptable values 0 1 Default value module dependent Description For each dimension it is possible to define the bound ary condition 0 and 1 stands for periodic and repulsive boundary conditions respectively e inerti
17. represent a class Classes highlighted in blue are common to any optimization problem and can be considered as a black box for by the user Classes in the yellow area change depending on the problem being solved We call a module a file containing a definition for these classes aimed at solving a specific problem Input is provided as a text file containing keywords with associated values and is managed by the Space class Converseley management of boundary conditions is the same for any optimization problem and is implemented in the DefaultSpace class The class PSO implements POW s optimizer The optimizer consists of a variation of a Particle Swarm Optimization algorithm called PSO Kick and Reseed PSO KaR The behavior of the optimizer is defined by an ensemble of parameters called inertia personal best cp neighborhood best cn and kar threshold kar_threshold Default values for these parameters are set the user is however free to set them at will using specific keywords in the parameterization file see section Standard Keywords Along the optimization run every measure performed by every particule is stored in a log file In order to extract useful information postprocessing this log file is necessary The class Postprocess is in charge of this Useful functions the user might need such as the selection of measures below a given threshold are preimplemented in the DefaultPostprocess class POW has been concieved so that th
18. s 50 particles and 3 repetitions has been chosen boundaryMin and boundary Max keyword de fine a multimer with a radius varying from 8 to 12 The provided monomer protein pdb will be treated as a rigid body and assembled in a heptameric structure 7 fold simmetry being constrained by constrain py function In postprocessing only solutions having a fitness smaller than 0 will be retained and solutions having an RMSD smaller than 5 within themselves will be clus tered By replacing the monomer keyword of previous example with what follows it s possible to perform a flexible assembly style flexible topology proten prmtop trajectory trajectory dcd align yes ratio 0 80 Flexible assembly requires a trajectory in crd or ded format and a topology pdb or psf If the protein in the trajectory is not aligned POW can do this for you by means of the align keyword This done PCA is performed on Cy atoms Notice that the number of degrees of freedom 3 N where N is the number of carbons must be greater than the number of frames in the sim ulation A number of eigenvectors representing more than 0 8 80 of the system s energy will be extracted and treated as protein s degrees of freedom Aligning the trajectory and performing a PCA may take some time However preprocessing phase will generate an aligned trajectory aligned dcd and a file containing eigenvectors projection proj_coordinates dat You can 20 indicate POW
19. ser e PSO py parallel implementation of Particle Swarm Optimization 4 Launching POW is launched in the console by means of the following command mpiexec n 4 INSTALLATION_PATH POW module input dat It is advised to create an alias in order to make POW execution easier The following lines create a default call using 4 processors export NPROC 4 alias pow mpiexec n NPROC POW_DIR POW py An execution becomes now as simple as pow module input dat This call will launch POW on 4 processors proper execution requires the user to provide two arguments to the call the desired optimization module module and a parameterization file input dat The parameterization file describes with a series of keywords how POW should behave The input file providing all the parametrisations for the search should be passed as param eter The file is structured as a serie of keywords one per line having one or more corresponding values Keywords are case sensitive and their order is irrelevant Some keywords are necessary for any kind of optimization proce dure see section whereas other are module specific see sections dedicated to specific modules The symbol can be used to comment out lines in parameterization file 5 Standard Keywords The following keywords implemented in Default py are typical to any optimization problem and are therefore accessible by any module e steps lt number of steps to per form gt Acceptable val
20. solutions gt Acceptable values float gt 0 Default value 5 Description Similar solutions will be clustered in a unique solution If RMSD clustering is chosen a value smaller or equal to 5 Ais adviced If distance clustering is used a number around 15 is suggested e output_ folder lt folder containing produced pdb structures gt Acceptable values string Default value result Description POW will generate a set of pdb corresponding to the clustering of best solutions These along with a summary file solu tions dat will be stored in the folder output_folder Note that the Default keywords boundaryMin and boundaryMax see Default keywords section 5 should include the following quantities in the following order see Search Space Definition and Data manipulation section 7 2 2 a B y radius 7 4 Constraint File The constraint file is user provided and contains a python function contain ing user defined measure on the generated multimer In the absence of a receptor this script consists of one function accepting a Multimer object that must be declared as follows def constraint_check multimer Huser defined measures return measurel measure2 In case a receptor is also present in the optimization process constraint_check will have to accept two parameters the second being the receptor Protein object The user can define various measures inside this function and return them The return order is significant it should
21. st be defined as well If flexible is chosen at least topology and trajectory keywords must be defined 15 monomer lt monomer PDB file gt Acceptable values UNIX filename Description PDB file containing the monomer Requires style key word set to rigid trajectory lt coordinates of a MD trajectory gt Acceptable values path to a ded or crd file Description Enesemble of protein structures Requires style keyword set to flexible topology lt topology of a MD trajectory gt Acceptable values path to a charmm or amber topology Description Topology of provided trajectory see trajectory key word Requires style keyword set to flexible trajSelection lt atom selection in MDAnalysis format gt Acceptable values MDAnalysis AtomSelect Default value protein Description Select a subset of atoms from provided trajectory If align keyword is set to yes trajectory will also be aligned on this se lection PCA and subsequent assembly will only take these atoms into account Requires style keyword set to flexible projection lt projection of MD trajectory on main eigenvectors gt Acceptable values path to a projections file Description If provided Principal Components Analysis will not be performed and this file providing projections on main eigenvectors will be used instead This file should consist of a number of lines matching the number of atoms in the provided trajectory and a number of columns correspondi
22. to use these file to avoid repeting the preprocessing This can be done in this way assembly_style flexible topology proten prmtop trajectory aligned dcd align no projection proj_coordinates dat 8 Creation of a new POW Module A module contains an implementation for Parser Data Space Fitness and Postprocess classes The following lines represent a module skeleton from Default import Parser as R from Default import Space as S from Default import Postprocess as PP import other packages here class Parser P def __init__ self infile parse more params if needed see Default py or DockSymmCircle py for syntax def check_variables self here you can perform consistency check on your parameters class Data def __init__ self params load files previously parsed contained in params object class Space S def __init__ self params data build search space using params and data objects defining self low low boundaries self high high boundaries self boundary_type int array 0 periodic 1 reflex 21 class Fitness def init__ self data params load data here if needed e g target measures def evaluate self num pos return fitness value class Postprocess PP def __init__ self params data load params and data structure def run self parse logfile and postprocess 22
23. ues positive integer Default value 100 Description The number of steps that will be computed in the PSO e particles lt number of particles gt Acceptable values positive integer Default value 40 Description The number of particles that will be used in each step of the PSO e repeat lt number of repetition gt Acceptable values positive integer Default value 1 Description Repeat can be used to lauch PSO multiple consecutive times This is useful in order to enhance the sampling e repulsion lt activate desactivate gt Acceptable values on off Default value off Description When repulsion is activated every good solution so lution smaller than filter_threshold particle velocity converging to zero found by PSO will be flagged Particles will be repelled by flagged regions with a x potential When performing multiple PSO repeti tions flags are passed from one PSO run to the following one This enhances PSO sampling since regions where a minima has been already discovered are not oversampled This option is currently experimental e neighborType lt type of neighbor gt Acceptable values indexed geographic Default value geographic Description NeighborType set the kind of neighborship between par ticles In indexed neighborhood particles are assigned an index and particles having consecutive indexes are considered as neighbots In geographic neighborhood distance within particles in the search s
24. vious log file will be backed up and a new one will be generated The new log file will contain all the data of previous logfile recordered until the restart point If the old log file is not found a new one is started Function Module The Function module allows the minimization of any function not requiring manipulation of any data structure The file containing the fitness function to be evaluated is passed to POW via the following keyword fitness lt fitness extraction file gt Acceptable values UNIX filename Default value fit_multimer Description This file contains the implementation for the Fitness class and should have the following form class Fitness def __init__ self data params pass def evaluate self num pos num PSO particle index pos array of particle s position in search space compute fitness on the base of pos values return fitness The Function module can also be operated via a graphical interface invoked with the command Function_GUI py see Figure 2 The interface allows the user to create edit and save a POW input file validate it and launch a POW run on multiple processors Notice that the use of this graphical interface also requires the wxPython package to be installed psoalbm JAG File 1 0 FILES Fitness file home backup PS0 src PSO_GUIv2 OFft_ select Log file og select BOUNDARY CONDITIONS ID
25. zation PSO search subsequently tries to arrange the elements of the conformational database in a multimeric assembly so that all restraints are respected and steric clashes avoided Geometric restraints can be typically provided by low resolution electron density maps or experiments such as cross linking disulfide scanning mutagenesis or FRET If necessary POW can assemble a multimer on a given substrate At PSO search com pletion a large set of solutions having a good score is usually generated A smaller set of representative solutions typically less than ten is returned by clustering the accepted solutions according to their respective Root Mean Square Deviation RMSD At present POW can predict hetero dimers when no symmetry is imposed i e addressing general protein protein interactions or homo multimers with or without a target substrate if a circular symmetry is defined This process is usually very fast less than 5 minutes on an average workstation using 4 processors and can produce small ensemble of solutions being sufficiently good to generate biologically sound working hypotheses and act as seeds for further optimization steps using more computationally expensive techniques 7 2 Implementation Details 7 2 1 Data Structures In order to manipulate protein structures two classes are implemented Protein and Multimer The Protein class allows to parse a PDB file manipulate its coordinates and extract the coordinates
Download Pdf Manuals
Related Search