RegGTFExtractor.py

#!/usr/bin/env python

"""

RegGTFExtractor.py extracts regulatory-data from Ensembl and UCSC databases
and converts the output to a GTF-formatted file.

@author: Sebastian Beyvers
@contact: sebastian.beyvers@med.uni-giessen.de

"""

import argparse
from Modules.Ensembl.Ensembl import Ensembl
from Modules.ucsc.ucsc import UcscGtf
from Modules.Uniquifier import UniqueFilter
from Modules.SaveResults import ResultSaver
from Modules.CrossMapper import CrossMapper
from Modules.Validator import Validator
import os
import json


def check_for_local_folder(wd):

    # Check if local folder exists and create if missing when no data_dir is specified
    # input_parameter: wd = working directory
    # return_value: None

    if not os.path.isdir(os.path.join(wd+"/data/")):

        os.mkdir(os.path.join(wd+"/data/"))

    if not os.path.isdir(os.path.join(wd+"/data/EnsemblData")):

        os.mkdir(os.path.join(wd+"/data/EnsemblData"))

    if not os.path.isdir(os.path.join(wd+"/data/UCSCData")):
        os.mkdir(os.path.join(wd+"/data/UCSCData"))

    if not os.path.isdir(os.path.join(wd+"/data/temp")):
        os.mkdir(os.path.join(wd+"/data/temp"))


def check_for_data_dir(data_dir):

    # Check if local folder exists and create if missing when data_dir as parameter is specified
    # input_parameter: data_dir = data directory
    # return_value: None

    if not os.path.isdir(os.path.join(data_dir)):
        os.mkdir(os.path.join(data_dir))

    if not os.path.isdir(os.path.join(data_dir+"/EnsemblData")):
        os.mkdir(os.path.join(data_dir+"/EnsemblData"))

    if not os.path.isdir(os.path.join(data_dir+"/UCSCData")):
        os.mkdir(os.path.join(data_dir + "/UCSCData"))

    if not os.path.isdir(os.path.join(data_dir+"/temp")):
        os.mkdir(os.path.join(data_dir+"/temp"))


def check_filter(tissue_cmd, org, wd):

    # Checks if filter-celltype is in Json types for organism
    # input_parameter: tissue_cmd: Filtered tissuetypes; org = organism; wd = working directory
    # return_value: boolean if selected filter is in config

    path_to_config = os.path.join(wd + "/config/celltypes_" + org + ".json")
    tissues_config = []
    if not tissue_cmd:
        return False
    with open(path_to_config) as input_file:
        data = json.loads(input_file.read())
        for x in data:
            tissues_config.append(x["type"])

    if any(tissue in tissues_config for tissue in tissue_cmd):
        return True

    else:
        return False


def check_organism(org):

    # Checks the organism input and decides if chrossmapping is necessary
    # input_parameter: org = input organism (parameter)
    # return_value: tuple with values = (organism_alias (string), boolean if chrossmapping is needed)

    if org == "hg38":
        return "homo_sapiens", False
    if org == "hg19":
        print("Older assembly Version detected: hg19 -> Crossmapping result from hg38")
        return "homo_sapiens", True
    elif org == "mm10":
        return "mus_musculus", False
    elif org == "mm9":
        print("Older assembly Version detected: mm9 -> Crossmapping result from mm10")
        return "mus_musculus", True


def main_script(organism, wd, data_dir, out, tissuetype=None):

    # main function
    # input_parameter: all parameters from argparse

    # if no output parameter is given output file is "./organism.gtf"
    if out == ".":
        out = "./"+organism+".gtf"

    (org,  x_mappable) = check_organism(organism)
    if not data_dir:
        check_for_local_folder(wd)
    else:
        check_for_data_dir(data_dir)
    if check_filter(tissuetype, org, wd):
        tissues = tissuetype
        print("Filter detected !")
    else:
        tissues = None
        print("Filter not detected !")

    # Get UCSC Data
    ucsc = UcscGtf(org, wd, data_dir)
    # Gen Ensembl Data
    ense = Ensembl(org, wd, data_dir)
    print("Getting Unique Results")
    unique_filter = UniqueFilter(ense.get_gtf(), ucsc.get_gtf(), tissues)
    if data_dir:
        ResultSaver(unique_filter.get_results(), organism, data_dir, x_mappable, True, out)
        if x_mappable:
            CrossMapper(organism, data_dir, out, True)
    else:
        ResultSaver(unique_filter.get_results(), organism, wd, x_mappable, False, out)
        if x_mappable:
            CrossMapper(organism, wd, out, False)

    # Validate outputfile

    Validator(out)


if __name__ == '__main__':

    # argument parser

    parser = argparse.ArgumentParser(description='GTF-Generator from UCSC Table Browser and Ensembl Regulatory Build' )
    parser.add_argument('organism', help='Source organism [ hg19 | hg38 or mm9 | mm10 ]', action='store', nargs='?', type=str)
    parser.add_argument('--tissue', help='Tissue- or Celltype(s)', action='store', nargs='*', type=str)
    parser.add_argument('--wd', help='Working directory. default: "."', action='store', default=os.getcwd(), type=str)
    parser.add_argument('--dir', help='Data directory. default: "working_directory"', action='store', default="", type=str)
    parser.add_argument('--out', help='Path to output file: default: "./organism.gtf"', action='store', default=".", type=str)
    args = vars(parser.parse_args())

    # Check if organism exists

    if args["organism"]:
        if args["organism"] in ["hg19", "hg38", "mm9", "mm10"]:
            print("Working Dir: " + args["wd"])
            main_script(args["organism"], args["wd"], args["dir"],  args["out"], args["tissue"])
        else:
            print("Invalid Organism: " + args["organism"] + " see -h for help")
    else:
        print("No Arguments found -> See python3 ./RegGTFExtractor.py -h for help.")