kkonganti@11: #!/usr/bin/env python3
kkonganti@11: 
kkonganti@11: # Kranti Konganti
kkonganti@11: 
kkonganti@11: import os
kkonganti@11: import glob
kkonganti@11: import pickle
kkonganti@11: import argparse
kkonganti@11: import inspect
kkonganti@11: import logging
kkonganti@11: import re
kkonganti@11: import pprint
kkonganti@11: from collections import defaultdict
kkonganti@11: 
kkonganti@11: # Multiple inheritence for pretty printing of help text.
kkonganti@11: class MultiArgFormatClasses(argparse.RawTextHelpFormatter, argparse.ArgumentDefaultsHelpFormatter):
kkonganti@11:     pass
kkonganti@11: 
kkonganti@11: 
kkonganti@11: # Main
kkonganti@11: def main() -> None:
kkonganti@11:     """
kkonganti@11:     This script works only in the context of `bettercallsal` Nextflow workflow.
kkonganti@11:     It takes:
kkonganti@11:         1. A pickle file containing a dictionary object where genome accession
kkonganti@11:             is the key and the computed serotype is the value.
kkonganti@11:         2. A file with `mash screen` results run against the Salmonella SNP
kkonganti@11:             Cluster genomes' sketch.
kkonganti@11:         3. A directory containing genomes' FASTA in gzipped format where the
kkonganti@11:             FASTA file contains 2 lines: one FASTA header followed by
kkonganti@11:             genome Sequence.
kkonganti@11:     and then generates a concatenated FASTA file of top N unique `mash screen`
kkonganti@11:     genome hits as requested.
kkonganti@11:     """
kkonganti@11: 
kkonganti@11:     # Set logging.
kkonganti@11:     logging.basicConfig(
kkonganti@11:         format="\n" + "=" * 55 + "\n%(asctime)s - %(levelname)s\n" + "=" * 55 + "\n%(message)s\n\n",
kkonganti@11:         level=logging.DEBUG,
kkonganti@11:     )
kkonganti@11: 
kkonganti@11:     # Debug print.
kkonganti@11:     ppp = pprint.PrettyPrinter(width=55)
kkonganti@11:     prog_name = os.path.basename(inspect.stack()[0].filename)
kkonganti@11: 
kkonganti@11:     parser = argparse.ArgumentParser(
kkonganti@11:         prog=prog_name, description=main.__doc__, formatter_class=MultiArgFormatClasses
kkonganti@11:     )
kkonganti@11: 
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-s",
kkonganti@11:         dest="sero_snp_metadata",
kkonganti@11:         default=False,
kkonganti@11:         required=False,
kkonganti@11:         help="Absolute UNIX path to metadata text file with the field separator, | "
kkonganti@11:         + "\nand 5 fields: serotype|asm_lvl|asm_url|snp_cluster_id"
kkonganti@11:         + "\nEx: serotype=Derby,antigen_formula=4:f,g:-|Scaffold|402440|ftp://...\n|PDS000096654.2\n"
kkonganti@11:         + "Mentioning this option will create a pickle file for the\nprovided metadata and exits.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-fs",
kkonganti@11:         dest="force_write_pick",
kkonganti@11:         action="store_true",
kkonganti@11:         required=False,
kkonganti@11:         help="By default, when -s flag is on, the pickle file named *.ACC2SERO.pickle\n"
kkonganti@11:         + "is written to CWD. If the file exists, the program will not overwrite\n"
kkonganti@11:         + "and exit. Use -fs option to overwrite.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-m",
kkonganti@11:         dest="mash_screen_res",
kkonganti@11:         default=False,
kkonganti@11:         required=False,
kkonganti@11:         help="Absolute UNIX path to `mash screen` results file.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-ms",
kkonganti@11:         dest="mash_screen_res_suffix",
kkonganti@11:         default=".screened",
kkonganti@11:         required=False,
kkonganti@11:         help="Suffix of the `mash screen` result file.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-ps",
kkonganti@11:         dest="pickled_sero",
kkonganti@11:         default=False,
kkonganti@11:         required=False,
kkonganti@11:         help="Absolute UNIX Path to serialized metadata object in a pickle file.\n"
kkonganti@11:         + "You can create the pickle file of the metadata using -s option.\n"
kkonganti@11:         + "Required if -m is on.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-gd",
kkonganti@11:         dest="genomes_dir",
kkonganti@11:         default=False,
kkonganti@11:         required=False,
kkonganti@11:         help="Absolute UNIX path to a directory containing\n"
kkonganti@11:         + "gzipped genome FASTA files.\n"
kkonganti@11:         + "Required if -m is on.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-gds",
kkonganti@11:         dest="genomes_dir_suffix",
kkonganti@11:         default="_scaffolded_genomic.fna.gz",
kkonganti@11:         required=False,
kkonganti@11:         help="Genome FASTA file suffix to search for\nin the directory mentioned using\n-gd.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-n",
kkonganti@11:         dest="num_uniq_hits",
kkonganti@11:         default=10,
kkonganti@11:         help="This many number of serotype genomes' accessions are " + "\nreturned.",
kkonganti@11:     )
kkonganti@11:     parser.add_argument(
kkonganti@11:         "-op",
kkonganti@11:         dest="out_prefix",
kkonganti@11:         default="MASH_SCREEN",
kkonganti@11:         help="Set the output file prefix for .fna.gz and .txt files.",
kkonganti@11:     )
kkonganti@11:     # required = parser.add_argument_group('required arguments')
kkonganti@11: 
kkonganti@11:     args = parser.parse_args()
kkonganti@11:     num_uniq_hits = int(args.num_uniq_hits)
kkonganti@11:     mash_screen_res = args.mash_screen_res
kkonganti@11:     mash_screen_res_suffix = args.mash_screen_res_suffix
kkonganti@11:     pickle_sero = args.sero_snp_metadata
kkonganti@11:     pickled_sero = args.pickled_sero
kkonganti@11:     f_write_pick = args.force_write_pick
kkonganti@11:     genomes_dir = args.genomes_dir
kkonganti@11:     genomes_dir_suffix = args.genomes_dir_suffix
kkonganti@11:     out_prefix = args.out_prefix
kkonganti@11:     mash_genomes_gz = os.path.join(
kkonganti@11:         os.getcwd(), out_prefix + "_TOP_" + str(num_uniq_hits) + "_UNIQUE_HITS.fna.gz"
kkonganti@11:     )
kkonganti@11:     mash_uniq_hits_txt = os.path.join(
kkonganti@11:         os.getcwd(), re.sub(".fna.gz", ".txt", os.path.basename(mash_genomes_gz))
kkonganti@11:     )
kkonganti@11: 
kkonganti@11:     if mash_screen_res and os.path.exists(mash_genomes_gz):
kkonganti@11:         logging.error(
kkonganti@11:             "A concatenated genome FASTA file,\n"
kkonganti@11:             + f"{os.path.basename(mash_genomes_gz)} already exists in:\n"
kkonganti@11:             + f"{os.getcwd()}\n"
kkonganti@11:             + "Please remove or move it as we will not "
kkonganti@11:             + "overwrite it."
kkonganti@11:         )
kkonganti@11:         exit(1)
kkonganti@11: 
kkonganti@11:     if os.path.exists(mash_uniq_hits_txt) and os.path.getsize(mash_uniq_hits_txt) > 0:
kkonganti@11:         os.remove(mash_uniq_hits_txt)
kkonganti@11: 
kkonganti@11:     if mash_screen_res and (not genomes_dir or not pickled_sero):
kkonganti@11:         logging.error("When -m is on, -ps and -gd are also required.")
kkonganti@11:         exit(1)
kkonganti@11: 
kkonganti@11:     if genomes_dir:
kkonganti@11:         if not os.path.isdir(genomes_dir):
kkonganti@11:             logging.error("UNIX path\n" + f"{genomes_dir}\n" + "does not exist!")
kkonganti@11:             exit(1)
kkonganti@11:         if len(glob.glob(os.path.join(genomes_dir, "*" + genomes_dir_suffix))) <= 0:
kkonganti@11:             logging.error(
kkonganti@11:                 "Genomes directory"
kkonganti@11:                 + f"{genomes_dir}"
kkonganti@11:                 + "\ndoes not seem to have any\n"
kkonganti@11:                 + f"files ending with suffix: {genomes_dir_suffix}"
kkonganti@11:             )
kkonganti@11:             exit(1)
kkonganti@11: 
kkonganti@11:     if pickle_sero and os.path.exists(pickle_sero) and os.path.getsize(pickle_sero) > 0:
kkonganti@11:         acc2serotype = defaultdict()
kkonganti@11:         init_pickled_sero = os.path.join(os.getcwd(), out_prefix + ".ACC2SERO.pickle")
kkonganti@11: 
kkonganti@11:         if (
kkonganti@11:             os.path.exists(init_pickled_sero)
kkonganti@11:             and os.path.getsize(init_pickled_sero)
kkonganti@11:             and not f_write_pick
kkonganti@11:         ):
kkonganti@11:             logging.error(
kkonganti@11:                 f"File {os.path.basename(init_pickled_sero)} already exists in\n{os.getcwd()}\n"
kkonganti@11:                 + "Use -fs to force overwrite it."
kkonganti@11:             )
kkonganti@11:             exit(1)
kkonganti@11: 
kkonganti@11:         with open(pickle_sero, "r") as sero_snp_meta:
kkonganti@11:             for line in sero_snp_meta:
kkonganti@11:                 cols = line.strip().split("|")
kkonganti@11:                 url_cols = cols[3].split("/")
kkonganti@11: 
kkonganti@11:                 if not 4 <= len(cols) <= 5:
kkonganti@11:                     logging.error(
kkonganti@11:                         f"The metadata file {pickle_sero} is malformed.\n"
kkonganti@11:                         + f"Expected 4-5 columns. Got {len(cols)} columns.\n"
kkonganti@11:                     )
kkonganti@11:                     exit(1)
kkonganti@11: 
kkonganti@11:                 if not len(url_cols) > 5:
kkonganti@11:                     acc = url_cols[3]
kkonganti@11:                 else:
kkonganti@11:                     acc = url_cols[9]
kkonganti@11: 
kkonganti@11:                 if not re.match(r"^GC[AF]\_\d+\.\d+$", acc):
kkonganti@11:                     logging.error(
kkonganti@11:                         f"Did not find accession in either field number 4\n"
kkonganti@11:                         + "or field number 10 of column 4."
kkonganti@11:                     )
kkonganti@11:                     exit(1)
kkonganti@11: 
kkonganti@11:                 acc2serotype[acc] = cols[0]
kkonganti@11: 
kkonganti@11:         with open(init_pickled_sero, "wb") as write_pickled_sero:
kkonganti@11:             pickle.dump(file=write_pickled_sero, obj=acc2serotype)
kkonganti@11: 
kkonganti@11:         logging.info(
kkonganti@11:             f"Created the pickle file for\n{os.path.basename(pickle_sero)}.\n"
kkonganti@11:             + "This was the only requested function."
kkonganti@11:         )
kkonganti@11:         sero_snp_meta.close()
kkonganti@11:         write_pickled_sero.close()
kkonganti@11:         exit(0)
kkonganti@11:     elif pickle_sero and not (os.path.exists(pickle_sero) and os.path.getsize(pickle_sero) > 0):
kkonganti@11: 
kkonganti@11:         logging.error(
kkonganti@11:             "Requested to create pickle from metadata, but\n"
kkonganti@11:             + f"the file, {os.path.basename(pickle_sero)} is empty or\ndoes not exist!"
kkonganti@11:         )
kkonganti@11:         exit(1)
kkonganti@11: 
kkonganti@11:     if mash_screen_res and os.path.exists(mash_screen_res):
kkonganti@11:         if os.path.getsize(mash_screen_res) > 0:
kkonganti@11: 
kkonganti@11:             seen_uniq_hits = 0
kkonganti@11:             unpickled_acc2serotype = pickle.load(file=open(pickled_sero, "rb"))
kkonganti@11: 
kkonganti@11:             with open(mash_screen_res, "r") as msh_res:
kkonganti@11:                 mash_hits = defaultdict()
kkonganti@11:                 seen_mash_sero = defaultdict()
kkonganti@11: 
kkonganti@11:                 for line in msh_res:
kkonganti@11:                     cols = line.strip().split("\t")
kkonganti@11: 
kkonganti@11:                     if len(cols) < 5:
kkonganti@11:                         logging.error(
kkonganti@11:                             f"The file {os.path.basename(mash_screen_res)} seems to\n"
kkonganti@11:                             + "be malformed. It contains less than required 5-6 columns."
kkonganti@11:                         )
kkonganti@11:                         exit(1)
kkonganti@11: 
kkonganti@11:                     mash_hit_acc = re.sub(
kkonganti@11:                         genomes_dir_suffix,
kkonganti@11:                         "",
kkonganti@11:                         str((re.search(r"GC[AF].*?" + genomes_dir_suffix, cols[4])).group()),
kkonganti@11:                     )
kkonganti@11: 
kkonganti@11:                     if mash_hit_acc:
kkonganti@11:                         mash_hits.setdefault(cols[0], []).append(mash_hit_acc)
kkonganti@11:                     else:
kkonganti@11:                         logging.error(
kkonganti@11:                             "Did not find an assembly accession in column\n"
kkonganti@11:                             + f"number 5. Found {cols[4]} instead. Cannot proceed!"
kkonganti@11:                         )
kkonganti@11:                         exit(1)
kkonganti@11:             msh_res.close()
kkonganti@11:         elif os.path.getsize(mash_screen_res) == 0:
kkonganti@11:             failed_sample_name = os.path.basename(mash_screen_res).rstrip(mash_screen_res_suffix)
kkonganti@11:             with open(
kkonganti@11:                 os.path.join(os.getcwd(), "_".join([out_prefix, "FAILED.txt"])), "w"
kkonganti@11:             ) as failed_sample_fh:
kkonganti@11:                 failed_sample_fh.write(f"{failed_sample_name}\n")
kkonganti@11:                 failed_sample_fh.close()
kkonganti@11:             exit(0)
kkonganti@11: 
kkonganti@11:         # ppp.pprint(mash_hits)
kkonganti@11:         msh_out_txt = open(mash_uniq_hits_txt, "w")
kkonganti@11:         with open(mash_genomes_gz, "wb") as msh_out_gz:
kkonganti@11:             for _, (ident, acc_list) in enumerate(sorted(mash_hits.items(), reverse=True)):
kkonganti@11:                 for acc in acc_list:
kkonganti@11:                     if seen_uniq_hits >= num_uniq_hits:
kkonganti@11:                         break
kkonganti@11:                     if unpickled_acc2serotype[acc] not in seen_mash_sero.keys():
kkonganti@11:                         seen_mash_sero[unpickled_acc2serotype[acc]] = 1
kkonganti@11:                         seen_uniq_hits += 1
kkonganti@11:                         # print(acc.strip() + '\t' + ident + '\t' + unpickled_acc2serotype[acc], file=sys.stdout)
kkonganti@11:                         msh_out_txt.write(
kkonganti@11:                             f"{acc.strip()}\t{unpickled_acc2serotype[acc]}\t{ident}\n"
kkonganti@11:                         )
kkonganti@11:                         with open(
kkonganti@11:                             os.path.join(genomes_dir, acc + genomes_dir_suffix), "rb"
kkonganti@11:                         ) as msh_in_gz:
kkonganti@11:                             msh_out_gz.writelines(msh_in_gz.readlines())
kkonganti@11:                         msh_in_gz.close()
kkonganti@11:         msh_out_gz.close()
kkonganti@11:         msh_out_txt.close()
kkonganti@11:         logging.info(
kkonganti@11:             f"File {os.path.basename(mash_genomes_gz)}\n"
kkonganti@11:             + f"written in:\n{os.getcwd()}\nDone! Bye!"
kkonganti@11:         )
kkonganti@11:         exit(0)
kkonganti@11: 
kkonganti@11: 
kkonganti@11: if __name__ == "__main__":
kkonganti@11:     main()