build_pop.py

#!/usr/bin/env python3

import sys
import os
import random
import argparse
from collections import OrderedDict
import vcf
from Bio import SeqIO
import tempfile
from lib.svrunner_utils import eprint
from pysam import tabix_compress, tabix_index


def parse_args():
    """
    Parse script arguments
    :return: argparse arguments object
    """
    parser = argparse.ArgumentParser(description='Generate simulated populations with SV')
    parser.add_argument("--nb-inds", help="Number of individuals to generate", required=True, type=int)
    parser.add_argument("--reference", help="Reference genome", required=True)
    parser.add_argument("--sv-list", help="File containing the SVs", required=True)
    parser.add_argument("--coverage", help="Coverage of reads (default: 15)", default=15, type=int)
    parser.add_argument("--output-directory", help="Output directory (default: res)", default="res")
    parser.add_argument("--force-polymorphism", help="Force polymorphism for each SV", action='store_const', const=True,
                        default=False)
    parser.add_argument("--haploid", help="Make a haploid genome, instead of diploid one", action="store_const", const=True,
                        default=False)
    parser.add_argument("-l", "--read-len", help="Generate reads having a length of LEN", type=int, default=100)
    parser.add_argument("-m", "--insert-len-mean", help="Generate inserts (fragments) having an average length of LEN",
                        type=int, default=300)
    parser.add_argument("-v", "--insert-len-sd", help="Set the standard deviation of the insert (fragment) length (%%)",
                        type=int, default=30)

    args = parser.parse_args()
    return args


def _allele(frequency):
    """
    Get randomly an allele, given a frequency
    :param frequency: frequency of the allele {float}
    :return: allele randomly choosen {0 or 1}
    """
    return 1 if random.uniform(0, 1) < frequency else 0


def _get_genotypes_for_inds(nb_inds, haploid, freq):
    """
    Get genotypes for each individual
    :param nb_inds: number of individuals {int}
    :param haploid: is the genome hamploid {bool}
    :param freq: frequency of the allele {float}
    :return: list of genotypes, vcf data {list}
    """
    all_genotypes = []
    genotypes_row = []
    for i in range(1, nb_inds + 1):
        if not haploid:
            genotype = str(_allele(freq)) + "/" + str(_allele(freq))
        else:
            genotype = str(_allele(freq))
        genotype_data = vcf.model._Call(None, "INDIV_" + str(i), vcf.model.make_calldata_tuple("GT")(GT=genotype))
        genotypes_row.append(genotype_data)
        all_genotypes.append(genotype)
    return all_genotypes, genotypes_row


def _svsort(sv, chrm, genotypes_for_inds):
    """
    Function to sort regions
    :param sv: the variant
    :param chrm: chromosome {str}
    :param genotypes_for_inds: dictionary of genotypes for each individual
    """
    return int(genotypes_for_inds[chrm][sv]["start"])


###########################
# 1. Get random deletions #
###########################

def get_random_deletions(sv_list, reference, tmp_dir, prg_path):
    """
    Get random deletions
    :param sv_list: file describing deletions (in SVsim format) {str}
    :param reference: reference genome fasta file {str}
    :param tmp_dir: temporary directory {str}
    :param prg_path: program path {str}
    """

    print("GENERATE SV DELs...\n")

    os.system(os.path.join(prg_path, "SVsim -W -i {0} -r {1} -o {2}{3}reference-sv".format(sv_list, reference, tmp_dir, os.path.sep)))


###################################
# 2. Build BED files of deletions #
###################################

def build_bed_deletions_file(bed_file, tmp_dir):
    """
    Build BED file of deletions
    :param bed_file: bed full path name {str}
    :param tmp_dir: temporary directory {str}
    """
    try:
        with open(bed_file, "w") as bed:
            try:
                with open(os.path.join(tmp_dir, "reference-sv.bedpe"), "r") as bedpe:
                    for line in bedpe:
                        freq = 0.2 if random.uniform(0,1) < 0.5 else 0.5
                        parts = line.split("\t")
                        bed.write("\t".join([parts[0], parts[2], parts[4], parts[6].replace("::", "_"), str(freq)]) + "\n")
            except IOError:
                eprint("ERROR: SVSim failed to generate variants.")
                exit(1)
    except IOError:
        eprint("ERROR: Unable to generate BED file \"{0}\".".format(os.path.join(tmp_dir, "reference-sv.bed")))
        exit(1)


###############################################################################
# 3. Build VCF files containing genotypes for the given number of individuals #
###############################################################################
# VCF file is a result file, not used by this script

def _build_vcf_header(vcf_file, prg_path, tmp_dir, nb_inds):
    """
    Build header of the VCF file
    :param vcf_file: vcf file full path name {str}
    :param prg_path: program path {str}
    :param tmp_dir: temporary directory {str}
    :param nb_inds: number of individuals {int}
    """
    try:
        with open(os.path.join(prg_path, "template.vcf"), "r") as template:
            try:
                with open(vcf_file, "w") as my_template:
                    for line in template:
                        if line[:6] == "#CHROM":
                            line = line.replace("\n", "")
                            for i in range(0, nb_inds):
                                line += "\tINDIV_" + str(i+1)
                            line += "\n"
                        my_template.write(line)
            except IOError:
                eprint("ERROR: unable to create template file \"{0}\" in temp dir.".format(os.path.join(tmp_dir, "template.vcf")))
                exit(1)
    except IOError:
        eprint("ERROR: template file not found in program directory.")
        exit(1)


def build_genotypes_vcf_list(bed_file, output_vcf, haploid, force_polymorphism, nb_inds, tmp_dir, prg_path):
    """
    Build VCF file describing genotypes for each individual (and the associated python dictionary)
    :param bed_file: bed file describing deletions {str}
    :param output_vcf: output VCF file full path name {str}
    :param haploid: is haploid {bool}
    :param force_polymorphism: force polymorphism {bool}
    :param output_dir: output directory {str}
    :param nb_inds: number of individuals {int}
    :param tmp_dir: temporary directory {str}
    :param prg_path: program path {str}
    :return: dictionary of genotypes for each variant for each dictionary {OrderedDict}
    """
    print("GENERATE SUMMARY VCF FILE...\n")
    genotypes_for_inds = OrderedDict()
    # { chr: { id_indiv: {start: #start, end: #end, genotypes: [[0,1],[1,1],...], ...}, # ...}

    vcf_file = os.path.join(tmp_dir, "template.vcf")

    # Build VCF header:
    _build_vcf_header(vcf_file, prg_path, tmp_dir, nb_inds)

    try:
        with open(bed_file, "r") as bed:
            vcf_reader = vcf.Reader(filename=vcf_file)
            vcf_writer = vcf.Writer(open(output_vcf, "w"), vcf_reader)
            for line in bed:
                parts = line.replace("\n", "").split("\t")
                freq = float(parts[4])
                if parts[0] not in genotypes_for_inds:
                    genotypes_for_inds[parts[0]] = {}
                genotypes_for_inds[parts[0]][parts[3]] = {"start": int(parts[1]), "end": int(parts[2]), "genotypes": []}

                # Get genotypes:
                all_genotypes, genotypes = [], []
                if force_polymorphism:
                    polymorph = False
                    while not polymorph:
                        all_genotypes, genotypes = _get_genotypes_for_inds(nb_inds, haploid, freq)
                        polymorph = len(set(all_genotypes)) > 1
                else:
                    all_genotypes, genotypes = _get_genotypes_for_inds(nb_inds, haploid, freq)
                genotypes_for_inds[parts[0]][parts[3]]["genotypes"] = [x.split("/") for x in all_genotypes]

                info = {"END": int(parts[2]), "AF": freq}
                vcf_record = vcf.model._Record(parts[0], int(parts[1]), parts[3], "N", [vcf.model._SV("DEL")], ".", ".", info, "GT", [0], genotypes)
                vcf_writer.write_record(vcf_record)
            vcf_writer.close()

            tabix_compress(output_vcf, output_vcf + ".gz", True)
            tabix_index(output_vcf + ".gz", force=True, preset="vcf")
    except IOError:
        eprint("ERROR: Unable to load bed file \"{0}\": file not found.".format(bed_file))
        exit(1)

    return genotypes_for_inds


###############################################
# Build fasta chromosomes for each individual #
###############################################

def _compute_keeped_genomic_regions(svs, svs_infos, haploid):
    """
    Get list of all regions keeped (not deleted)
    :param svs: list of variants (deletions) {list}
    :param svs_infos: infos of variants {dict}
    :param haploid: is haploid {bool}
    :return: regions for each individuals {OrderedDict}, the last position for each region {OrderedDict}
    """
    regions = OrderedDict()
    current_region_pointer = OrderedDict()
    for svs_i in svs:
        i = 0
        for genotypes in svs_infos[svs_i]["genotypes"]:
            if i not in regions:
                regions[i] = {}
                current_region_pointer[i] = {}
            for j in range(0, 2 if not haploid else 1):  # For each chromosome of the diploid genome, or the chromosome
                # of the haploid genome
                if j not in regions[i]:
                    regions[i][j] = []
                    current_region_pointer[i][j] = "0"
                if svs_infos[svs_i]["genotypes"][i][j] == "1":
                    regions[i][j].append([current_region_pointer[i][j], svs_infos[svs_i]["start"]])
                    current_region_pointer[i][j] = svs_infos[svs_i]["end"]
            i += 1
    return regions, current_region_pointer


def _build_fastas(chrm, regions, current_region_pointer, output_dir, fasta_orig_chr, last_nt):
    """
    Build fasta files
    :param chrm:
    :param regions:
    :param current_region_pointer:
    :param output_dir:
    :param fasta_orig_chr:
    :param last_nt:
    :return:
    """
    for indiv, chrs_dips in regions.items():
        id_chrs = list(chrs_dips.keys())
        id_chrs = sorted(id_chrs)
        for id_chr in id_chrs:
            chr_dip = chrs_dips[id_chr]  # SVs for each diploid chromosome
            logs_regions = []  # Store logs

            # Build FASTA and store logs:
            try:
                with open(os.path.join(output_dir, "INDIV_" + str(indiv + 1) + "_chr_" + str(id_chr) + ".fasta"),
                          "a") as output_handle:
                    fasta = ""
                    last_one = 0
                    for chr_region in chr_dip:
                        fasta += fasta_orig_chr[int(chr_region[0]):int(chr_region[1])]
                        logs_regions.append("\t".join(str(x) for x in chr_region))
                        last_one = int(chr_region[1])
                    if last_one < last_nt:
                        logs_regions.append("\t".join([str(current_region_pointer[indiv][id_chr]), str(last_nt)]))
                        fasta += fasta_orig_chr[int(current_region_pointer[indiv][id_chr]):last_nt]
                    SeqIO.write(fasta, output_handle, "fasta")
            except IOError:
                eprint("ERROR: unable to write \"{0}\" file.".
                       format(os.path.join(output_dir, "INDIV_" + str(indiv + 1) + "_chr_" + str(id_chr) + ".fasta")))
                exit(1)

            # Write logs
            try:
                with open(os.path.join(output_dir, "INDIV_" + str(indiv + 1) + ".regions.log"), "a") as log_handle:
                    log_handle.write(chrm + "_" + str(id_chr) + "\t")
                    log_handle.write("\n\t".join(logs_regions))
                    log_handle.write("\n")
            except IOError:
                eprint("ERROR: unable to write log file: \"{0}\"".
                       format(os.path.join(output_dir, "INDIV_" + str(indiv + 1) + ".regions.log")))


def build_fastas_chromosomes(reference, genotypes_for_inds, haploid, output_dir):
    print("BUILD FASTA GENOME FOR EACH INDIVIDUAL...\n")
    fasta_orig = SeqIO.index(reference, "fasta")

    for chrm, svs_infos in genotypes_for_inds.items():

        print("PROCESSING CHROMOSOME {0}...\n".format(chrm))

        svs = list(svs_infos.keys())
        svs = sorted(svs, key=lambda x:_svsort(x, chrm, genotypes_for_inds))

        fasta_orig_chr = fasta_orig[chrm]
        last_nt = len(fasta_orig_chr)-1

        # Compute keeped genomic regions for each diploid chromosome:
        regions, current_region_pointer = _compute_keeped_genomic_regions(svs, svs_infos, haploid)

        # Build FASTA of each diploid/haploid chromosome:
        _build_fastas(chrm, regions, current_region_pointer, output_dir, fasta_orig_chr, last_nt)


def generate_samples_fastq(haploid, nb_inds, output_dir, coverage, read_len, insert_len_mean, insert_len_sd,
                           prg_path):

    print("GENERATE RANDOM READS FOR EACH INDIVIDUAL FROM GENOME...\n")

    # Generate reads for all individuals:
    if not haploid:
        cmd = str("{4}pirs/pirs simulate -z -x {3} -d -B {4}pirs/Profiles/Base-Calling_Profiles/humNew.PE100.matrix.gz "
                  "-I {4}pirs/Profiles/InDel_Profiles/phixv2.InDel.matrix -l {5} -m {6} -v {7} "
                  "-G {4}pirs/Profiles/GC-depth_Profiles/humNew.gcdep_100.dat -o {0} {1} {2}")
    else:
        cmd = str("{4}pirs/pirs simulate -z -x {3} -B {4}pirs/Profiles/Base-Calling_Profiles/humNew.PE100.matrix.gz "
                  "-I {4}pirs/Profiles/InDel_Profiles/phixv2.InDel.matrix -l {5} -m {6} -v {7} "
                  "-G {4}pirs/Profiles/GC-depth_Profiles/humNew.gcdep_100.dat -o {0} {1}")
    for i in range(1, nb_inds+1):
        prefix = os.path.join(output_dir, "INDIV_" + str(i))
        chr0 = prefix + "_chr_0.fasta"
        chr1 = prefix + "_chr_1.fasta"
        if not haploid:
            os.system(cmd.format(prefix, chr0, chr1, coverage, prg_path + os.path.sep, read_len,
                                 insert_len_mean, insert_len_sd))
        else:
            os.system(cmd.format(prefix, chr0, "", coverage, prg_path + os.path.sep, read_len,
                                 insert_len_mean, insert_len_sd))


def main():
    args = parse_args()
    reference = args.reference
    sv_list = args.sv_list
    output_dir = args.output_directory
    tmp_dir = tempfile.mkdtemp()
    haploid = args.haploid
    nb_inds = args.nb_inds

    if os.path.isdir(output_dir):
        eprint("Error: output directory {0} already exists.".format(output_dir))
        return 1
    elif os.path.isfile(output_dir):
        eprint("Error: unable to create output directory {0}: file exists.".format(output_dir))
        return 1

    if nb_inds < 2:
        raise Exception("nb-inds must be at least 2")

    if not os.path.isfile(reference + ".fai"):
        os.system("samtools faidx " + reference)

    prg_path = os.path.dirname(os.path.realpath(__file__))

    os.mkdir(output_dir)

    ############################
    # Define fixed files names #
    ############################
    bed_file = os.path.join(tmp_dir, "reference-sv.bed")
    output_vcf = os.path.join(output_dir, "genotypes.vcf")

    ################
    # Launch steps #
    ################

    get_random_deletions(sv_list, reference, tmp_dir, prg_path)
    build_bed_deletions_file(bed_file, tmp_dir)
    genotypes_for_inds = build_genotypes_vcf_list(bed_file, output_vcf, haploid, args.force_polymorphism, nb_inds, tmp_dir, prg_path)
    build_fastas_chromosomes(reference, genotypes_for_inds, haploid, output_dir)
    generate_samples_fastq(haploid, nb_inds, output_dir, args.coverage, args.read_len, args.insert_len_mean,
                           args.insert_len_sd, prg_path)
    print("DONE!\n")

if __name__ == '__main__':
    sys.exit(main())