GTFGen.py

import os
import gzip


class GTFGen:

    """

    Class to generate Ensembl GTF-data with activity
    @author: Sebastian Beyvers
    @contact: sebastian.beyvers@med.uni-giessen.de

    """

    def __init__(self, organism, release, wd, data_dir):

        # Constructor for GTFGen
        # input_parameter: organism = input organism
        #                  release  = used Ensembl release
        #                  wd       = working directory (default is "."), this is used if data_dir is not specified.
        #                  data_dir = data directory (if specified this is used)

        self.gff_lines = self.get_organism_as_gff(organism, release, wd, data_dir)

        # Map to assign numbers from Activitytable-binary to activity status

        self.value_map = {0: "ACTIVE", 1: "POISED", 2: "REPRESSED", 3: "INACTIVE", 4: "NA"}

    def get_organism_as_gff(self, organism, release, wd, data_dir):

        # reads the original gff file for organism
        # input_parameter as in __init__ described.
        # return_value: list of gff-entries

        if data_dir:
            directory = os.path.join(data_dir + "/EnsemblData/", release, organism)
        else:
            directory = os.path.join(wd + "/data/EnsemblData/", release, organism)
        inputfile = ""
        for file in os.listdir(directory):
            if file.endswith("gff.gz"):
                inputfile = os.path.join(directory, file)

        with gzip.open(inputfile) as original_file:
            return original_file.readlines()

    def reformat_to_gtf(self, activity, release):

        # Reformats gff to gtf and appends activity-data for config specified celltype-categories
        # input_parameter: activity = list of activity status for all genes
        #                  release = current ensembl release
        # return_value: List of gtf-formatted entries

        gtf_return = []

        for index, line in enumerate(self.gff_lines):

            decoded_line = line.decode("UTF-8")
            # Generate a templist
            templist = []
            # Split Line by Tab
            splitted = decoded_line.split("\t")
            # Split Last Field by ";"
            splitted_additional = splitted[-1].strip().split(";")
            # Add Chromosome Name Format = chr+Name
            templist.append("chr"+splitted[0])
            # Add RegBuild_ + release
            templist.append("RegBuild_"+release)
            # Add Description from Description in last ; separated segment
            templist.append(splitted_additional[4].split("=")[1].lower().replace(' ', '_'))
            # Add Start / End Data from original
            templist.extend(splitted[3:5])
            # Add Score, Strand and Frame Data
            templist.extend([".", "+", "."])
            # Add "additional" information

            templist.append(self.generate_additional_information(splitted_additional[0],
                            self.generate_activity_list(activity, index)))

            gtf_return.append(templist)

        return gtf_return

    @staticmethod
    def generate_additional_information(gene_id, activity):

        # helper method to concat activity information to string and reformat from gff to gtf-style
        # input_parameter: gene_id = gene_id formatted in gff format
        #                  activity = List of activity-data for specified gene
        # return_value: String for attributes (column 9) in gtf-format

        if gene_id.startswith("ID=regulatory_region:"):
            gene_id = 'gene_id "'+gene_id.split(':')[1]+'"'
        elif gene_id.startswith("ID=E"):
            gene_id = 'gene_id "'+gene_id.split('=')[1]+'"'

        activity_string = 'activity "'+', '.join(activity)+'"'

        return gene_id+'; '+activity_string+';'

    def generate_activity_list(self, activity, index):

        # generates activity list for a specified index
        # input_parameter: index = index for a specified gene
        #                  activity = List of activity-data for all entries
        # return_value: List of activity for gene at index

        activity_list = []
        for key, value in activity.items():
            activity_list.append(key+">"+self.value_map[value[index]])
        return activity_list

    def get_gtf(self, release, activity):

        # getter function for the resulting gtf-formatted-list
        # input_parameters: release, activity as in self.reformat_to_gtf()
        # return_value: List of GTF-Entries

        return self.reformat_to_gtf(activity, release)