view draw_amr_matrix.py @ 17:eb86b4bf140e draft default tip

Uploaded
author greg
date Wed, 03 May 2023 15:13:12 +0000
parents 7bd48449ee79
children
line wrap: on
line source

#!/usr/bin/env python

import argparse
import csv
import os
import subprocess
import sys
import tempfile

import Bio.SeqIO
import numpy
import pandas
import matplotlib.pyplot as pyplot


def load_fasta(fasta_file):
    sequence = pandas.Series(dtype=object)
    for contig in Bio.SeqIO.parse(fasta_file, 'fasta'):
        sequence[contig.id] = contig
    return sequence


def run_command(cmd):
    try:
        tmp_name = tempfile.NamedTemporaryFile(dir=".").name
        tmp_stderr = open(tmp_name, 'wb')
        proc = subprocess.Popen(args=cmd, shell=True, stderr=tmp_stderr.fileno())
        returncode = proc.wait()
        tmp_stderr.close()
        if returncode != 0:
            # Get stderr, allowing for case where it's very large.
            tmp_stderr = open(tmp_name, 'rb')
            stderr = ''
            buffsize = 1048576
            try:
                while True:
                    stderr += tmp_stderr.read(buffsize)
                    if not stderr or len(stderr) % buffsize != 0:
                        break
            except OverflowError:
                pass
            tmp_stderr.close()
            os.remove(tmp_name)
            stop_err(stderr)
    except Exception as e:
        stop_err('Command:\n%s\n\nended with error:\n%s\n\n' % (cmd, str(e)))


def stop_err(msg):
    sys.stderr.write(msg)
    sys.exit(1)


def draw_amr_matrix(amr_feature_hits_files, amr_deletions_file, varscan_vcf_file, amr_mutation_regions_bed_file, amr_gene_drug_file, reference, reference_size, mutation_regions_dir, amr_matrix_png_dir, errors):
    efh = open(errors, 'w')
    ofh = open('process_log', 'w')

    # Read amr_feature_hits_files.
    amr_feature_hits = pandas.Series(dtype=object)
    for amr_feature_hits_file in amr_feature_hits_files:
        ofh.write("\namr_feature_hits_file: %s\n" % amr_feature_hits_file)
        feature_name = os.path.basename(amr_feature_hits_file)
        ofh.write("\nfeature_name: %s\n" % feature_name)
        # Make sure the file is not empty.
        if os.path.isfile(amr_feature_hits_file) and os.path.getsize(amr_feature_hits_file) > 0:
            best_hits = pandas.read_csv(filepath_or_buffer=amr_feature_hits_file, sep='\t', header=None)
            ofh.write("\nFeature file %s will be processed\n" % os.path.basename(amr_feature_hits_file))
        else:
            ofh.write("\nEmpty feature file %s will NOT be processed\n" % os.path.basename(amr_feature_hits_file))
            best_hits = None
        amr_feature_hits[feature_name] = best_hits

    # Process populated feature hits.
    for k in amr_feature_hits.keys():
        if amr_feature_hits[k] is not None:
            amr_hits = amr_feature_hits[k]
            ofh.write("\namr_hits:\n%s\n" % str(amr_hits))
            amr_to_draw = pandas.DataFrame(columns=['gene', 'drug'])
            ofh.write("\namr_to_draw:\n%s\n" % str(amr_to_draw))
            # Read amr_drug_gene_file.
            amr_gene_drug = pandas.read_csv(amr_gene_drug_file, index_col=None, sep='\t', quoting=csv.QUOTE_NONE, header=None)
            ofh.write("\namr_gene_drug:\n%s\n" % str(amr_gene_drug))

            # Roll up AMR gene hits.
            ofh.write("\namr_hits.shape[0]: %s\n" % str(amr_hits.shape[0]))
            if amr_hits.shape[0] > 0:
                for gene_idx, gene in amr_hits.iterrows():
                    ofh.write("gene_idx: %s\n" % str(gene_idx))
                    ofh.write("gene: %s\n" % str(gene))
                    gene_name = gene[3]
                    ofh.write("gene_name: %s\n" % str(gene_name))
                    ofh.write("amr_gene_drug[0]: %s\n" % str(amr_gene_drug[0]))
                    drugs = amr_gene_drug.loc[amr_gene_drug[0] == gene_name, :][1]
                    ofh.write("drugs: %s\n" % str(drugs))
                    for drug in drugs:
                        amr_to_draw = amr_to_draw.append(pandas.Series([gene_name, drug], name=amr_to_draw.shape[0], index=amr_to_draw.columns))
                        ofh.write("\amr_to_draw: %s\n" % str(amr_to_draw))

            ofh.write("\nvarscan_vcf_file is None: %s\n" % str(varscan_vcf_file == 'None'))
            if varscan_vcf_file not in [None, 'None'] and os.path.getsize(varscan_vcf_file) > 0:
                amr_mutations = pandas.Series(dtype=object)
                if amr_mutation_regions_bed_file is not None:
                    mutation_regions = pandas.read_csv(amr_mutation_regions_bed_file, header=0, sep='\t', index_col=False)
                    # Validate mutation regions.
                    if mutation_regions.shape[1] != 7:
                        efh.write("The selected mutations regions BED file is invalid, it should be a six column file.\n")
                    elif mutation_regions.shape[0] == 0:
                        efh.write("There are no rows in the selected mutation regions file.\n")
                    else:
                        for region_i in range(mutation_regions.shape[0]):
                            region = mutation_regions.iloc[region_i, :]
                            if region[0] not in reference:
                                efh.write("Mutation region '%s' not found in reference genome.\n" % str(region))
                                break
                            if not isinstance(region[1], numpy.int64):
                                efh.write("Non-integer found in mutation region start (column 2): %s.\n" % str(region[1]))
                                break
                            if not isinstance(region[2], numpy.int64):
                                efh.write("Non-integer found in mutation region start (column 3): %s.\n" % str(region[2]))
                                break
                            if region[1] <= 0 or region[2] <= 0:
                                efh.write("Mutation region '%s' starts before the reference sequence.\n" % str(region))
                            if region[1] > len(reference[region[0]].seq) or region[2] > len(reference[region[0]].seq):
                                efh.write("Mutation region '%s' ends after the reference sequence.\n" % str(region))
                            if not region.get('type', default='No Type') in ['snp', 'small-indel', 'any']:
                                ofh.write("\n\nSkipping mutation region '%s' with invalid type '%s', valid types are 'snp', 'small-indel', 'any'.\n\n" % (str(region), str(region.get('type', default='No Type'))))
                                continue
                            ofh.write("\nFinding AMR mutations for %s.\n" % str(region['name']))
                            region_bed = 'region_%s.bed' % region_i
                            ofh.write("region_bed: %s\n" % str(region_bed))
                            mutation_regions.loc[[region_i], ].to_csv(path_or_buf=region_bed, sep='\t', header=False, index=False)
                            ofh.write("mutation_regions.loc[[region_i], ]:\n%s\n" % str(mutation_regions.loc[[region_i], ]))
                            region_mutations_tsv = os.path.join(mutation_regions_dir, 'region_%s_mutations.tsv' % region_i)
                            ofh.write("region_mutations_tsv: %s\n" % str(region_mutations_tsv))
                            cmd = ' '.join(['bedtools intersect',
                                            '-nonamecheck',
                                            '-wb',
                                            '-a', region_bed,
                                            '-b', varscan_vcf_file,
                                            ' | awk \'BEGIN{getline < "' + amr_mutation_regions_bed_file + '";printf $0"\\t";',
                                            'getline < "' + varscan_vcf_file + '"; getline < "' + varscan_vcf_file + '";print $0}{print}\'',
                                            '1>' + region_mutations_tsv])
                            ofh.write("\ncmd:\n%s\n" % cmd)
                            run_command(cmd)
                            try:
                                ofh.write("After running command, os.path.getsize((region_mutations_tsv): %s\n" % str(os.path.getsize(region_mutations_tsv)))
                                region_mutations = pandas.read_csv(region_mutations_tsv, sep='\t', header=0, index_col=False)
                                ofh.write("\nregion_mutations: %s\n" % region_mutations)
                            except Exception:
                                continue
                            # Figure out what kind of mutations are in this region.
                            region_mutation_types = pandas.Series(['snp'] * region_mutations.shape[0], name='TYPE', index=region_mutations.index)
                            region_mutation_types[region_mutations['REF'].str.len() != region_mutations['ALT'].str.len()] = 'small-indel'
                            region_mutation_drugs = pandas.Series(region['drug'] * region_mutations.shape[0], name='DRUG', index=region_mutations.index)
                            region_notes = pandas.Series(region['note'] * region_mutations.shape[0], name='NOTE', index=region_mutations.index)
                            region_mutations = pandas.concat([region_mutations, region_mutation_types, region_mutation_drugs, region_notes], axis=1)
                            region_mutations = region_mutations[['#CHROM', 'POS', 'TYPE', 'REF', 'ALT', 'DRUG', 'NOTE']]
                            amr_mutations[region['name']] = region_mutations
                else:
                    ofh.write("\nMutation region BED file not received.\n")
                ofh.write("\nAfter processing mutations, amr_mutations: %s\n" % str(amr_mutations))
                # Roll up potentially resistance conferring mutations.
                ofh.write("\n##### Rolling up potentially resistance conferring mutations..\n")
                for mutation_region, mutation_hits in amr_mutations.iteritems():
                    ofh.write("mutation_region: %s\n" % str(mutation_region))
                    ofh.write("mutation_hits: %s\n" % str(mutation_hits))
                    for mutation_idx, mutation_hit in mutation_hits.iterrows():
                        ofh.write("mutation_idx: %s\n" % str(mutation_idx))
                        ofh.write("mutation_hit: %s\n" % str(mutation_hit))
                        mutation_name = mutation_region + ' ' + mutation_hit['REF'] + '->' + mutation_hit['ALT']
                        ofh.write("mutation_name: %s\n" % str(mutation_name))
                        amr_to_draw = amr_to_draw.append(pandas.Series([mutation_name, mutation_hit['DRUG']], name=amr_to_draw.shape[0], index=amr_to_draw.columns))
                ofh.write("\nAfter processing mutations, amr_to_draw: %s\n" % str(amr_to_draw))
                ofh.write("\nAfter processing mutations, amr_to_draw.shape[0]: %s\n" % str(amr_to_draw.shape[0]))

            if amr_deletions_file not in [None, 'None'] and os.path.getsize(amr_deletions_file) > 0:
                # Roll up deletions that might confer resistance.
                try:
                    amr_deletions = pandas.read_csv(filepath_or_buffer=amr_deletions_file, sep='\t', header=None)
                except Exception:
                    amr_deletions = pandas.DataFrame()
                if amr_deletions.shape[0] > 0:
                    amr_deletions.columns = ['contig', 'start', 'stop', 'name', 'type', 'drug', 'note']
                    amr_deletions = amr_deletions.loc[amr_deletions['type'].isin(['large-deletion', 'any']), :]
                    for deletion_idx, deleted_gene in amr_deletions.iterrows():
                        amr_to_draw = amr_to_draw.append(pandas.Series(['\u0394' + deleted_gene[3], deleted_gene[5]], name=amr_to_draw.shape[0], index=amr_to_draw.columns))
                ofh.write("\nAfter processing deletions, amr_to_draw: %s\n" % str(amr_to_draw))

            ofh.write("\namr_to_draw.shape[0]: %s\n" % str(amr_to_draw.shape[0]))
            # I have no idea why, but when running functional tests with planemo
            # the value of amr_to_draw.shape[0] is 1 even though the tests use the
            # exact inputs when running outside of planeo that result in the value
            # being 2.  So we cannot test with planemo unless we incorporate a hack
            # like a hidden in_test_mode parameter.
            if amr_to_draw.shape[0] > 1:
                ofh.write("\nDrawing AMR matrix...\n")
                present_genes = amr_to_draw['gene'].unique()
                present_drugs = amr_to_draw['drug'].unique()
                amr_matrix = pandas.DataFrame(0, index=present_genes, columns=present_drugs)
                for hit_idx, hit in amr_to_draw.iterrows():
                    amr_matrix.loc[hit[0], hit[1]] = 1
                amr_matrix_png = os.path.join(amr_matrix_png_dir, 'amr_matrix.png')
                int_matrix = amr_matrix[amr_matrix.columns].astype(int)
                figure, axis = pyplot.subplots()
                heatmap = axis.pcolor(int_matrix, cmap=pyplot.cm.Blues, linewidth=0)
                axis.invert_yaxis()
                axis.set_yticks(numpy.arange(0.5, len(amr_matrix.index)), minor=False)
                axis.set_yticklabels(int_matrix.index.values)
                axis.set_xticks(numpy.arange(0.5, len(amr_matrix.columns)), minor=False)
                axis.set_xticklabels(amr_matrix.columns.values, rotation=90)
                axis.xaxis.tick_top()
                axis.xaxis.set_label_position('top')
                pyplot.tight_layout()
                pyplot.savefig(amr_matrix_png, dpi=300)
            else:
                ofh.write("\nEmpty AMR matrix, nothing to draw...\n")
    efh.close()
    ofh.close()


if __name__ == '__main__':
    parser = argparse.ArgumentParser()

    parser.add_argument('--amr_feature_hits_dir', action='store', dest='amr_feature_hits_dir', help='Directory of tabular files containing feature hits')
    parser.add_argument('--amr_deletions_file', action='store', dest='amr_deletions_file', default=None, help='AMR deletions BED file')
    parser.add_argument('--varscan_vcf_file', action='store', dest='varscan_vcf_file', default=None, help='Varscan VCF file produced by the call_amr_mutations tool')
    parser.add_argument('--amr_mutation_regions_bed_file', action='store', dest='amr_mutation_regions_bed_file', default=None, help='AMR mutation regions BED file')
    parser.add_argument('--amr_gene_drug_file', action='store', dest='amr_gene_drug_file', help='AMR_gene_drugs tsv file')
    parser.add_argument('--reference_genome', action='store', dest='reference_genome', help='Reference genome fasta file')
    parser.add_argument('--mutation_regions_dir', action='store', dest='mutation_regions_dir', help='Directory for mutation regions TSV files produced by this tool')
    parser.add_argument('--amr_matrix_png_dir', action='store', dest='amr_matrix_png_dir', help='Directory for PNG files produced by this tool')
    parser.add_argument('--errors', action='store', dest='errors', help='Output file containing errors')

    args = parser.parse_args()

    # Get the collection of feature hits files.  The collection
    # will be sorted alphabetically and will contain 2 files
    # named something like AMR_CDS_311_2022_12_20.fasta and
    # Incompatibility_Groups_2023_01_01.fasta.
    amr_feature_hits_files = []
    for file_name in sorted(os.listdir(args.amr_feature_hits_dir)):
        file_path = os.path.abspath(os.path.join(args.amr_feature_hits_dir, file_name))
        amr_feature_hits_files.append(file_path)

    # Load the reference genome into memory.
    reference = load_fasta(args.reference_genome)
    reference_size = 0
    for i in reference:
        reference_size += len(i.seq)

    draw_amr_matrix(amr_feature_hits_files, args.amr_deletions_file, args.varscan_vcf_file, args.amr_mutation_regions_bed_file, args.amr_gene_drug_file, reference, reference_size, args.mutation_regions_dir, args.amr_matrix_png_dir, args.errors)