Download pgen

Of the script demands the overhead of processing a model. To compute the pgens of many sequences, it is suggested to read the sequences in from a file and use either mode 2 or 3.It is also possible to condition the Pgen computation on V and J identity by specifying the V or J usages as a mask. However, note that these V/J masks will be applied to ALL of the sequences provided as arguments. Read Options on v_mask and j_mask for more info.Example calls:Compute the pgen of an amino acid CDR3 sequence$ olga-compute_pgen --humanTRB CASSTGQANYGYTFPgen of the amino acid sequence CASSTGQANYGYTF: 5.26507446955e-08Compute the pgen of an in-frame nucleotide sequence and the amino acid sequence it translates to.$ olga-compute_pgen --humanTRB TGTGCCAGCAGTGACGCACAGGGGCGTAATCGTGGGACTGAAGCTTTCTTTPgen of the nucleotide sequence TGTGCCAGCAGTGACGCACAGGGGCGTAATCGTGGGACTGAAGCTTTCTTT: 1.31873701121e-17Pgen of the amino acid sequence nt2aa(TGTGCCAGCAGTGACGCACAGGGGCGTAATCGTGGGACTGAAGCTTTCTTT) = CASSDAQGRNRGTEAFF: 4.70599549953e-13Compute the pgen of a regular expression template of CDR3 amino acid sequences. Note, for a regular expression sequence, provided as an argument, backslashes may be needed to specify the characters {} for the sequence to be read in properly.$ olga-compute_pgen --humanTRB CASSTGX\{1,5\}QAN[YA]GYTFPgen of the regular expression sequence CASSTGX{1,5}QAN[YA]GYTF: 7.588241802e-08Compute the pgens of all three sequences.$ olga-compute_pgen --humanTRB CASSTGQANYGYTF CASSTGX\{1,5\}QAN[YA]GYTF TGTGCCAGCAGTGACGCACAGGGGCGTAATCGTGGGACTGAAGCTTTCTTTPgen of the amino acid sequence CASSTGQANYGYTF: 5.26507446955e-08Pgen of the regular expression sequence CASSTGX{1,5}QAN[YA]GYTF: 7.588241802e-08Pgen of the nucleotide sequence TGTGCCAGCAGTGACGCACAGGGGCGTAATCGTGGGACTGAAGCTTTCTTT: 1.31873701121e-17Pgen of the amino acid sequence nt2aa(TGTGCCAGCAGTGACGCACAGGGGCGTAATCGTGGGACTGAAGCTTTCTTT) = CASSDAQGRNRGTEAFF: 4.70599549953e-13Specify a comma delimited V or J mask to condition the pgen computation on V and/or J gene usage.$ olga-compute_pgen --humanTRB CASSTGQANYGYTF --v_mask TRBV2,TRBV14 --j_mask TRBJ1-2Pgen of the amino acid sequence CASSTGQANYGYTF: 1.39165562898e-09It is also possible to restrict the Pgen computation to specified V and/or J genes or alleles (to reflect any alignment outside of the CDR3 region) by using the options -v or -j (see example below). You can specify multiple V or J genes/alleles by using a comma as a delimiter.The only required inputs are the sequence and specifying the generative V(D)J model. Additional options can be found by using -h.Modes 2/3):These read in sequences from a file. The script has only minimal file parsing built in, so reading in sequences from a file

Requires the file to be structured with delimiter spaced values (i.e. the data is organized in columns separated by delimiter like a .tsv or .csv file). Read Options on delimiter for more info.To read in sequences, the index of column of CDR3 sequences is needed. The default is to assume that the sequences to be read in are in the first column (index 0), meaning that a text file with only a sequence on each line will be read in okay by default. Read Options on seq_in for more info.It is not recommended to read in regular expression sequences from a file. These sequences require enumerating out the amino acid sequences which correspond to them and computing pgen for each of them individually -- this can require a large time cost. Instead consider defining a custom 'amino acid' alphabet to define the symbols used in the regular expressions if possible. Furthermore, BE CAREFUL if reading in from a .csv file -- if commas are used in a regex sequence and comma is used as the delimiter of the .csv file, the sequence will not be read in properly.If nucleotide sequences are to be read in it is possible to specify if theoutput should be the nucleotide sequence Pgen and/or the translated amino acid sequence Pgen (the default is to compute and output both). See Options.It is also possible to condition the Pgen computation on V and J identity by specifying what index the column that V and J masks are stored for each line.Mode 2 does not have a specified output file and so will print the sequences and their pgens to stdout.Mode 3 does have a specified output file. By default in this mode there is a running display of the last few sequences/pgens written to the output file as well as time elapsed, current rate of computation, and estimated time remaining. This display can be disabled (see Options).As it is rare for datasets to have >> 1e4 unique sequences, parallelization is not built in to compute_pgen. However, there are options to skip N lines of the file and

And VJ recombination models, however the methods that get called are the same.The modules are:Module nameClassesload_model.pyGenomicDataVDJ, GenomicDataVJ, GenerativeModelVDJ, GenerativeModelVJpreprocess_generative_model_and_data.pyPreprocessedParametersVDJ, PreProcessedParametersVJgeneration_probability.pyGenerationProbabilityVDJ, GenerationProbabilityVJsequence_generation.pySequenceGenerationVDJ, SequenceGenerationVJutils.pyN/A (contains util functions)The classes with methods that are of interest will be GenerationProbabilityV(D)J (to compute Pgens) and SequenceGenerationV(D)J (to generate sequences).There is a fair amount of parameter processing that must go on to call these methods, however this is generally all done by instantiating a particular class. An exception to this rule are the classes GenerativeModelV(D)J and GenomicDataV(D)J. Normally the genomic data and model parameters are read in from IGoR inference files (and prepared V and J anchor files that have been prepared), however this is not mandated in order to make it easier for people to adapt the code to read in models/genomic data from other sources.Instantiating GenerativeModelV(D)J and GenomicDataV(D)J leaves the attributes as dummies, and calling the methods load_and_process_igor_model and load_igor_genomic_data will load up IGoR files.If you want to load models/data from other sources, you will need to write your own methods to set the attributes in GenerativeModelV(D)J and GenomicDataV(D)J. Please see the documentation of load_model.py for more details.Here is an example of loading the default human TRB model to compute some sequence Pgens and to generate some random CDR3 sequences:>> import olga.load_model as load_model>>> import olga.generation_probability as pgen>>> import olga.sequence_generation as seq_gen>>>>>> #Define the files for loading in generative model/data... params_file_name = 'default_models/human_T_beta/model_params.txt'>>> marginals_file_name = 'default_models/human_T_beta/model_marginals.txt'>>> V_anchor_pos_file ='default_models/human_T_beta/V_gene_CDR3_anchors.csv'>>> J_anchor_pos_file = 'default_models/human_T_beta/J_gene_CDR3_anchors.csv'>>>>>> #Load data... genomic_data = load_model.GenomicDataVDJ()>>> genomic_data.load_igor_genomic_data(params_file_name, V_anchor_pos_file, J_anchor_pos_file)>>> #Load model... generative_model = load_model.GenerativeModelVDJ()>>> generative_model.load_and_process_igor_model(marginals_file_name)>>>>>> #Process model/data for pgen computation by instantiating GenerationProbabilityVDJ... pgen_model = pgen.GenerationProbabilityVDJ(generative_model, genomic_data)>>>>>> #Compute some sequence pgens... pgen_model.compute_regex_CDR3_template_pgen('CASSAX{0,5}SARPEQFF')6.846877804096558e-10>>> pgen_model.compute_aa_CDR3_pgen('CAWSVAPDRGGYTF', 'TRBV30*01', 'TRBJ1-2*01')1.203646865765782e-10>>> pgen_model.compute_nt_CDR3_pgen('TGTGCCAGTAGTATAACAACCCAGGGCTTGTACGAGCAGTACTTC')3.9945642868171824e-14>>>>>>>>> #Process model/data for sequence generation by instantiating SequenceGenerationVDJ... seq_gen_model = seq_gen.SequenceGenerationVDJ(generative_model, genomic_data)>>>>>> #Generate some random sequences... seq_gen_model.gen_rnd_prod_CDR3()('TGTGCCAGCAGTGAAAAAAGGCAATGGGAAAGCGGGGAGCTGTTTTTT', 'CASSEKRQWESGELFF', 27, 8)>>> seq_gen_model.gen_rnd_prod_CDR3()('TGTGCCAGCAGTTTAGTGGGAAGGGCGGGGCCCTATGGCTACACCTTC', 'CASSLVGRAGPYGYTF', 14, 1)>>> seq_gen_model.gen_rnd_prod_CDR3()('TGTGCCAGCTGGACAGGGGGCAACTACGAGCAGTACTTC', 'CASWTGGNYEQYF', 55, 13)">>>> import olga.load_model as load_model>>> import olga.generation_probability as pgen>>> import olga.sequence_generation as seq_gen>>>>>> #Define the files for loading in generative model/data... params_file_name = 'default_models/human_T_beta/model_params.txt'>>> marginals_file_name = 'default_models/human_T_beta/model_marginals.txt'>>> V_anchor_pos_file ='default_models/human_T_beta/V_gene_CDR3_anchors.csv'>>> J_anchor_pos_file = 'default_models/human_T_beta/J_gene_CDR3_anchors.csv'>>>>>> #Load data... genomic_data = load_model.GenomicDataVDJ()>>> genomic_data.load_igor_genomic_data(params_file_name, V_anchor_pos_file, J_anchor_pos_file)>>> #Load model... generative_model = load_model.GenerativeModelVDJ()>>> generative_model.load_and_process_igor_model(marginals_file_name)>>>>>> #Process model/data for pgen computation by