#!/usr/bin/env python

# Created by Christopher Rushton 2020/04/20

# This converts a MAF file and copy number matrix (optional) into the input files used be the LymphGen lymphoma

# classifier. See https://llmpp.nih.gov/lymphgen/index.php for more information

import argparse

import math

import os

import bisect

import logging

from collections import Counter

from sortedcontainers import SortedSet

class Gene:

    """

    A simple class storing the chromosome, start, end, and name of a gene

    """

    __slots__ = ["name", "chrom", "start", "end"]

    def __init__(self, chrom, start, end, name):

        self.name = name

        self.chrom = chrom

        self.start = start

        self.end = end

class Chromosome:

    """

    A simple class storying the genomic range of each chromosome, as well as the coordiates of the p and q arm

    """

    def __init__(self, chrom):

        self.chrom = chrom

        self.p_start = None

        self.p_end = None

        self.q_start = None

        self.q_end = None

        self.q_length = 0.0

        self.p_length = 0.0

    def add(self, start, end, arm):

        if arm == "p":

            # Check to see if we already have genomic coordinates for this arm

            if self.p_start is not None or self.p_end is not None:

                raise AttributeError("Coordinates already assigned for %s arm of chromosome \'%s\'" % (arm, self.chrom))

            self.p_start = start

            self.p_end = end

            self.p_length = end - start

            # Sanity check this does not overlap the q arm

            if self.q_start is not None and self.p_end > self.q_start:

                raise AttributeError("For chromosome \'%s\'. The q arm starts before the end of the p arm" % self.chrom)

        elif arm == "q":

            # Check to see if we already have genomic coordinates for this arm

            if self.q_start is not None or self.q_end is not None:

                raise AttributeError("Coordinates already assigned for %s arm of chromosome \'%s\'" % (arm, self.chrom))

            self.q_start = start

            self.q_end = end

            self.q_length = end - start

            # Sanity check this does not overlap the p arm

            if self.p_end is not None and self.q_start < self.p_end:

                raise AttributeError("For chromosome \'%s\'. The q arm starts before the end of the p arm" % self.chrom)

        else:

            # Not the p or q arm. Invalid

            raise AttributeError("Unknown arm specified for chromosome \'%s\':\'%s\'" % (self.chrom, arm))

class SampleCNVs:

    """

    Stores all the copy number events within a given class

    I was going to write this in a function, but having a class is a lot cleaner

    """

    def __init__(self):

        self.starts = {}

        self.ends = {}

        self.cn_states = {}

        self.len_seg = {}

        self.is_chr_prefixed = None

        self.ploidy = None

    def merge_overlapping_cnvs(self, existStarts: iter, existEnds: iter, existCNs: iter, newStart: int, newEnd: int, newCN:int):

        """

        Handle overlapping copy number segments, merging and editing existing segments as necessary

        In essence, if this segment overlaps an existing segment with the same copy number state, those segments are merged

        together. If the copy number state is different, the largest increase is kept,

         truncating or even breaking apart the copy-neutral segment in the process. In the worse case senario,

        this will lead to three different segments being created

        :param existStarts: The start position of the existing segment

        :param existEnds: The end position of the existing segment

        :param existCNs: The copy number state of the existing segment

        :param newStart: The start position of the new segment

        :param newEnd: The end position of the new segment

        :param newCN: The copy number state of the new segment

        :return: A tuple containing the new and old events. Ex {[newEvent1, newEvent2], [oldEvent1, oldEvent2]}

        """

        def reduce_new_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart, oldEnd):

            # The existing event has higher priority than the old event

            # Truncate the newer event

            # Since this could break the newer event apart into two pieces, lets do that now, and see

            # if the output segments are valid

            outStart1 = newStart

            outEnd1 = oldStart

            outStart2 = oldEnd

            outEnd2 = newEnd

            if outStart1 < outEnd1:  # Valid segment, process

                existStarts, existEnds, existCNs = self.merge_overlapping_cnvs(existStarts, existEnds, existCNs,

                                                                             outStart1, outEnd1, newCN)

            if outStart2 < outEnd2:  # Valid segment

                existStarts, existEnds, existCNs = self.merge_overlapping_cnvs(existStarts, existEnds, existCNs,

                                                                             outStart2, outEnd2, newCN)

            return existStarts, existEnds, existCNs

        def reduce_old_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart, oldEnd, oldCN):

            # The newer event is a "higher"-level deletion

            # Split/Truncate the older event

            outStart1 = oldStart

            outEnd1 = newStart

            outStart2 = newEnd

            outEnd2 = oldEnd

            outCN = oldCN

            # Delete existing event

            existStarts.pop(start_bisect)

            existEnds.pop(start_bisect)

            existCNs.pop(start_bisect)

            if outStart2 < outEnd2:  # Valid segment, do the last one first to maintain sorted order

                existStarts.insert(start_bisect, outStart2)

                existEnds.insert(start_bisect, outEnd2)

                existCNs.insert(start_bisect, outCN)

            if outStart1 < outEnd1:  # Also valid

                existStarts.insert(start_bisect, outStart1)

                existEnds.insert(start_bisect, outEnd1)

                existCNs.insert(start_bisect, outCN)

            # Check for any more overlaps

            return self.merge_overlapping_cnvs(existStarts, existEnds, existCNs, newStart, newEnd, newCN)

        # First, does this new segment actually overlap anything?

        start_bisect = bisect.bisect_right(existEnds, newStart)

        end_bisect = bisect.bisect_left(existStarts, newEnd)

        if start_bisect == end_bisect:

            # There are no overlaps. Simply add this new event in, and return it

            existStarts.insert(start_bisect, newStart)

            existEnds.insert(start_bisect, newEnd)

            existCNs.insert(start_bisect, newCN)

            return existStarts, existEnds, existCNs

        # Grab the first overlap

        oldStart = existStarts[start_bisect]

        oldEnd = existEnds[start_bisect]

        oldCN = existCNs[start_bisect]

        # Simplest case first. If both these events have the same CN state, lets merge them together

        if oldCN == newCN:

            outStart = oldStart if oldStart < newStart else newStart

            outEnd = oldEnd if oldEnd > newEnd else newEnd

            # Delete the existing event

            existStarts.pop(start_bisect)

            existEnds.pop(start_bisect)

            existCNs.pop(start_bisect)

            # Check for any more overlaps

            return self.merge_overlapping_cnvs(existStarts, existEnds, existCNs, outStart, outEnd, newCN)

        else:

            # These segments overlap and have different CN states.

            # Lets keep the highest-level event

            if newCN <= 2 and oldCN <= 2: # Deletion

                if oldCN < newCN:

                    # The older event is a "higher"-level deletion

                    return reduce_new_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart, oldEnd)

                else:

                    return reduce_old_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart, oldEnd, oldCN)

            if newCN >= 2 and oldCN >= 2:  # Gain/Amp

                if oldCN > newCN:

                    # The older event is a higher level gain. Split/truncate the new event

                    return reduce_new_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart,

                                            oldEnd)

                else:

                    return reduce_old_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart, oldEnd,

                                          oldCN)

            else:

                # One event must be a gain/amp, while the other is a deletion. In this case, keep both, and subset the

                # larger event by the smaller one

                if oldEnd - oldStart < newEnd - newStart:

                    # The older event is smaller. Split the newer event

                    return reduce_new_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart,

                                            oldEnd)

                else:

                    # The newer event is smaller. We should split the older event

                    return reduce_old_event(existStarts, existEnds, existCNs, newStart, newEnd, newCN, oldStart, oldEnd,

                                          oldCN)

    def add(self, chrom: str, start: int, end: int, cn: int):

        # For LymphGen compatability: Strip "chr" prefix

        chrom = chrom.replace("chr", "")

        if end <= start:

            if start == end:

                # This segment has a length of 0?!. Skip this I guesss

                return None

            else:

                # This segment has a negative length??

                raise TypeError("Invalid CNV segment: \'%s:%s-%s\'" % (chrom, start, end))

        # Have we seen any events for this chromosome before?

        if chrom not in self.len_seg:

            # If not, just store this segment. Simple!

            self.starts[chrom] = [start]

            self.ends[chrom] = [end]

            self.cn_states[chrom] = [cn]

            self.len_seg[chrom] = 1

            # Are we using chr-prefixed contig names?

            if self.is_chr_prefixed is None:

                self.is_chr_prefixed = True if chrom.startswith("chr") else False

        else:

            # If we have seen events on this chromosome before, we need to compare this new segment with existing segments

            # In theory, events should be non-overlapping, but I don't want to assume that because then things will break

            # horribly later

            self.starts[chrom], self.ends[chrom], self.cn_states[chrom] = \

                self.merge_overlapping_cnvs(self.starts[chrom], self.ends[chrom], self.cn_states[chrom], start, end, cn)

            self.len_seg[chrom] = len(self.cn_states)

    def overlap_chrom(self, chromosome: Chromosome, threshold: float = 0.8):

        """

        Calculates the overlap between the segments stored here and a given chromosome

        :param chromosome: A Chromosome() object defining the start and end of the p and q arms, respectively

        :return: A dictionary containing

        """

        # Do the CNVs within this chromosome encompass a given chromosome more than the specified threshold?

        chrom_name = chromosome.chrom

        # Handle "chr" prefix nonsense

        if self.is_chr_prefixed and not chrom_name.startswith("chr"):

            chrom_name = "chr" + chrom_name

        elif not self.is_chr_prefixed and chrom_name.startswith("chr"):

            chrom_name = chrom_name.replace("chr", "")

        try:

            chrom_starts = self.starts[chrom_name]

        except KeyError:

            # No CN events were provided for this chromosome, hence there can be no overlap

            return {}

        chrom_ends = self.ends[chrom_name]

        chrom_cn = self.cn_states[chrom_name]

        homdel_sum = {"p": 0, "q": 0, "Chrom": 0}

        del_sum = {"p": 0, "q": 0, "Chrom": 0}

        gain_sum = {"p": 0, "q": 0, "Chrom": 0}

        amp_sum = {"p": 0, "q": 0, "Chrom": 0}

        # Check for overlapping segments for each arm

        for (start, end, cn) in zip(chrom_starts, chrom_ends, chrom_cn):

            # Ignore copy-neutral segments

            if cn == 2:

                continue

            if end < chromosome.p_start or start > chromosome.q_end:

                # Segment falls out of range

                continue

            if start < chromosome.p_end:

                olap_start = start if start > chromosome.p_start else chromosome.p_start

                olap_end = end if end < chromosome.p_end else chromosome.p_end

                if cn < 2:

                    del_sum["p"] += olap_end - olap_start

                    del_sum["Chrom"] +=  olap_end - olap_start

                    if cn < 1:

                        homdel_sum["p"] += olap_end - olap_start

                        homdel_sum["Chrom"] += olap_end - olap_start

                elif cn > 2:

                    gain_sum["p"] += olap_end - olap_start

                    gain_sum["Chrom"] += olap_end - olap_start

                    if cn > 3:

                        amp_sum["p"] += olap_end - olap_start

                        amp_sum["Chrom"] += olap_end - olap_start

            if end > chromosome.q_start:  # We use an if, not elif, in case a segment overlaps both the p and q arm

                olap_start = start if start > chromosome.q_start else chromosome.q_start

                olap_end = end if end < chromosome.q_end else chromosome.q_end

                if cn < 2:

                    del_sum["q"] += olap_end - olap_start

                    del_sum["Chrom"] +=  olap_end - olap_start

                    if cn < 1:

                        homdel_sum["q"] += olap_end - olap_start

                        homdel_sum["Chrom"] += olap_end - olap_start

                elif cn > 2:

                    gain_sum["q"] += olap_end - olap_start

                    gain_sum["Chrom"] += olap_end - olap_start

                    if cn > 3:

                        amp_sum["q"] += olap_end - olap_start

                        amp_sum["Chrom"] += olap_end - olap_start

        events = {}

        # Now, calculate the fraction of each overlap

        # Start from the highest level and biggest event, and work our way down/smaller

        # Amplifications/gains

        if amp_sum["Chrom"] / (chromosome.p_length + chromosome.q_length) > threshold: # Whole chromosome Amp

            events[chrom_name + "Chrom"] = "AMP"

        # Now check for arm level events

        else:

            if amp_sum["p"] / chromosome.p_length > threshold:  # p Amp

                events[chrom_name + "p"] = "AMP"

            elif amp_sum["q"] / chromosome.q_length > threshold:  # q Amp

                events[chrom_name + "q"] = "AMP"

            if chrom_name + "Chrom" not in events and gain_sum["Chrom"] / (chromosome.p_length + chromosome.q_length) > threshold:  # Whole chromosome gain

                events[chrom_name + "Chrom"] = "GAIN"

            else:

                if chrom_name + "p" not in events and gain_sum["p"] / chromosome.p_length > threshold:  # p Gain

                    events[chrom_name + "p"] = "GAIN"

                if  chrom_name + "q" not in events and gain_sum["q"] / chromosome.q_length > threshold:  # q Gain

                    events[chrom_name + "q"] = "GAIN"

        # Homozygous and heterozygous deletions

        if homdel_sum["Chrom"] / (chromosome.p_length + chromosome.q_length) > threshold: # Whole chromosome Homozygous deletion

            events[chrom_name + "Chrom"] = "HOMDEL"

        # Now check for arm level events

        else:

            if homdel_sum["p"] / chromosome.p_length > threshold:  # p HOMDEL

                events[chrom_name + "p"] = "HOMDEL"

            elif homdel_sum["q"] / chromosome.q_length > threshold:  # q HOMDEL

                events[chrom_name + "q"] = "HOMDEL"

            if chrom_name + "Chrom" not in events and del_sum["Chrom"] / (chromosome.p_length + chromosome.q_length) > threshold:  # Whole chromosome deletions

                events[chrom_name + "Chrom"] = "HETLOSS"

            else:

                if chrom_name + "p" not in events and del_sum["p"] / chromosome.p_length > threshold:  # p deletion

                    events[chrom_name + "p"] = "HETLOSS"

                if  chrom_name + "q" not in events and del_sum["q"] / chromosome.q_length > threshold:  # q deletions

                    events[chrom_name + "q"] = "HETLOSS"

        return events

    def adjust_ploidy(self, arm_coords, sample_name, redo=False):

        """

        Adjust a sample's CN states based upon ploidy

        Calculate the average copy number state across the entire genome. If a sample is not diploid, normalize all the CN

        changes

        :param arm_coords: A dictionary containing {"chromosome_name": Chromosome()} objects which list chromosomal coordinates

        :param sample_name: A string specifying the sample name. Used for debugging/error purposes

        :param redo: Should we override any existing ploidy state for this sample?

        :return: None

        """

        if self.ploidy is not None and not redo:

            # Ploidy for this sample has already been calculated. Don't do anything

            return None

        # Store the length affected for each ploidy state

        ploidy_cov = {}

        genome_size = 0

        for chromosome in arm_coords.values():

            # Find overlapping CNVs for this arm

            # if the p or q arm coordinates are not set, use a placeholder

            if chromosome.p_start is None:

                chromosome.p_start = -100000

                chromosome.p_end = -100000

                chromosome.p_length = 1

            if chromosome.q_start is None:

                chromosome.q_start = -100000

                chromosome.q_end = -100000

                chromosome.q_length = 1

            # Save the size of this chromosome

            genome_size += chromosome.p_length

            genome_size += chromosome.q_length

            chrom_name = chromosome.chrom

            # Handle "chr" prefix nonsense

            if self.is_chr_prefixed and not chrom_name.startswith("chr"):

                chrom_name = "chr" + chrom_name

            elif not self.is_chr_prefixed and chrom_name.startswith("chr"):

                chrom_name = chrom_name.replace("chr", "")

            try:

                chrom_starts = self.starts[chrom_name]

            except KeyError:

                # No CN events were provided for this chromosome, hence there can be no overlap

                continue

            chrom_ends = self.ends[chrom_name]

            chrom_cn = self.cn_states[chrom_name]

            # Check for overlapping segments for each arm

            for (start, end, cn) in zip(chrom_starts, chrom_ends, chrom_cn):

                # Check p-arm overlap

                if end < chromosome.p_start or start > chromosome.q_end:

                    # Segment falls out of range

                    continue

                if start < chromosome.p_end:

                    olap_start = start if start > chromosome.p_start else chromosome.p_start

                    olap_end = end if end < chromosome.p_end else chromosome.p_end

                    # Store the length of this overlap and associated CN

                    if cn not in ploidy_cov:

                        ploidy_cov[cn] = 0

                    ploidy_cov[cn] += olap_end - olap_start

                if end > chromosome.q_start:  # We use an if, not elif, in case a segment overlaps both the p and q arm

                    olap_start = start if start > chromosome.q_start else chromosome.q_start

                    olap_end = end if end < chromosome.q_end else chromosome.q_end

                    # Store the length of this overlap and associated CN

                    if cn not in ploidy_cov:

                        ploidy_cov[cn] = 0

                    ploidy_cov[cn] += olap_end - olap_start

        # Now that we have calculated the number of bases affected by each CN state, calculate the ploidy

        x = 0

        for cn, bases in ploidy_cov.items():

            x += cn * bases

        ploidy = x / genome_size

        av_ploidy = round(ploidy)

        # Sanity check

        if av_ploidy < 1:

            raise TypeError("Ploidy of sample \'%s\' was calculated to be below 1!" % sample_name)

        # If this tumour is not diploid, adjust the ploidy

        if av_ploidy != 2:

            logging.debug("\'%s\' was calculated to have a ploidy of %s" % (sample_name, av_ploidy))

            ploidy_dif = av_ploidy - 2

            new_cns = {}

            for chrom, cn in self.cn_states.items():

                x = list(y - ploidy_dif for y in cn)

                new_cns[chrom] = x

            self.cn_states = new_cns

        self.ploidy = av_ploidy

def get_args():

    def is_valid_dir(path, parser):

        """

        Checks to ensure the directory path exists

        :param path: A string containing a filepath to a directory

        :param parser: An argparse.ArgumentParser() object

        :return: path, if the string is a valid directory

        :raises: parser.error() if the directory does not exist

        """

        if os.path.exists(path) and os.path.isdir(path):

            return path

        else:

            raise parser.error("Unable to set \'%s\' as the output directory: Not a valid directory" % path)

    def is_valid_file(path, parser):

        """

        Checks to ensure the specified file exists

        :param path: A string containing a filepath

        :param parser: An argparse.ArgumentParser() object

        :return: path, if the string is a valid directory

        :raises: parser.error() if the file does not exist

        """

        if os.path.exists(path) and os.path.isfile(path):

            return path

        else:

            raise parser.error("Unable to locate \'%s\': No such file or directory" % path)

    epilog = os.linesep.join(["The --arms, --entrez_ids, and --lymphgen_genes files can be found in the \'resources\' folder. ",

                              "Note that genome and exome sequencing types are handled exactly the same. ",

                              "The --entrez-ids file must have the Hugo_Symbol under the column \"Approved Symbol\" and the Entrez ID under the column \"NCBI Gene ID(supplied by NCBI)\". ",

                              "The --cnvs file should contain the following colummns: Tumor_Sample_Barcode, chromosome, start, end, CN. ",

                              "The --arms file should contain the following columns: chromosome, start, end, arm. "

                              ])

    parser = argparse.ArgumentParser(description="Generates input files for the LymphGen classifier\nVersion 1.0.0", epilog=epilog, formatter_class=argparse.RawDescriptionHelpFormatter)

    snvs_args = parser.add_argument_group("Input mutation files")

    snvs_args.add_argument("-m", "--maf", metavar="MAF", required=True, type=lambda x: is_valid_file(x, parser), help="Input MAF file listing somatic mutations")

    snvs_args.add_argument("-e", "--entrez_ids", metavar="TSV", required=True, type=lambda x: is_valid_file(x, parser), help="A tab-delimited file containing gene names (Hugo Symbol) and the corresponding Entrez gene ID")

    cnv_args = parser.add_argument_group("(Optional) Input CNV files")

    cnv_args.add_argument("-c", "--cnvs", metavar="TSV", default=None, type=lambda x: is_valid_file(x, parser), help="Input tab-delimited file summarizing copy number events")

    cnv_args.add_argument("--log2", action="store_true", help="Does the input CNV file use log2 ratios? (i.e. log2(absoluteCN) - 1)")

    cnv_args.add_argument("-g", "--genes", metavar="BED", default=None, type=lambda x: is_valid_file(x, parser), help="Input BED4+ file listing start and end positions of genes/exons")

    cnv_args.add_argument("-a", "--arms", metavar="TSV", default=None, type=lambda x: is_valid_file(x, parser), help="Input tab-delimited file listing the positions of chromosome arms")

    parser.add_argument("-l", "--lymphgen_genes", metavar="TXT", default=None, type=lambda x: is_valid_file(x, parser), required=False, help="An optional file listing all Entrez IDs considered by the LymphGen classifier. Output will be subset to these genes")

    parser.add_argument("-s", "--sequencing_type", metavar="SEQ", choices=["targeted", "exome", "genome"], required=True, help="Sequencing type used to obtain somatic mutations")

    parser.add_argument("-o", "--outdir", metavar="PATH", required=True, type=lambda x: is_valid_dir(x, parser), help="Output directory for LymphGen input files")

    parser.add_argument("--outprefix", metavar="STRING", default=None, help="Output files will be prefixed using this string [Default: Use the base name of the MAF file]")

    parser.add_argument("-v", "--verbose", choices=['DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL'], default="INFO", help="Set logging verbosity")

    args = parser.parse_args()

    # If no outprefix was set, use the basename of the maf file as the output prefix

    if args.outprefix is None:

        args.outprefix = os.path.basename(args.maf).split(".")[0]

    # If --cnvs is specified, the annotation files are also required

    if args.cnvs:

        if not args.genes or not args.arms:

            raise parser.error("--genes and --arms are required if --cnvs is specified")

    logging.basicConfig(level=getattr(logging, args.verbose), format='%(levelname)s:%(message)s')

    return args

def load_entrez_ids(entrez_file: str, hugo_column: str="Approved symbol", entrez_column: str="NCBI Gene ID(supplied by NCBI)",

                    status_column: str="Status", prev_name_column: str="Previous symbols"):

    """

    Creates a dictionary which contains the gene name (Hugo Symbol) and corresponding entrez ID (NCBI ID) from the

    specified text file

    Note the input file must have a header. An example file can be obtained from https://www.genenames.org/download/custom/

    Ensure the "Approved symbol" and "NCBI Gene ID" columns are selected. The default column names reflect those obtained

    from this source.

    The "Approval Status" column is optional, but will be used to remove gene names which have since been withdrawn.

    If the "Previous symbols" solumn is included, these gene names will also be assigned to that Entrez ID

    The subset_file should list one Entrez ID per line

    All extra columns are ignored

    :param entrez_file: A string containing a filepath to a tsv file containing the gene names and gene IDs

    :param hugo_column: A string specifying the name of the column which contains

    :param entrez_column: A string specifying the name of the column which contains the Entrez gene ID

    :param status_column: A string specifying the name of the column which contains the Approval Status. Optional.

    :param prev_name_column: A string specifting the name of the column which specifies previously used names for each gene. Optional

    :return: Two dictionaries containing {"Gene Name": "Entrez_ID"}

    """

    hugo_name_to_id = {}

    alt_hugo_name_to_id = {}  # If Previous symbols are included

    skipped_genes = []

    hugo_col_num = None

    entrez_col_num = None

    status_col_num = None  # Optional

    prev_name_col_num = None  # Optional

    i = 0

    with open(entrez_file) as f:

        for line in f:

            i += 1

            line = line.rstrip("\n").rstrip("\r")  # Remove line endings

            cols = line.split("\t")

            if len(cols) < 2:  # Sanity check this file has multiple columns

                raise TypeError("Input file %s does not appear to be a tab-delimited file" % entrez_file)

            # If we have not assigned column IDs yet, then we have not processed the file header

            # Load the file header

            if hugo_col_num is None or entrez_col_num is None:

                j = 0

                for col in cols:

                    if col == hugo_column:

                        hugo_col_num = j

                    if col == entrez_column:

                        entrez_col_num = j

                    if status_column is not None and col == status_column:

                        status_col_num = j

                    if prev_name_column is not None and col == prev_name_column:

                        prev_name_col_num = j

                    j += 1

                # Sanity check that we have found all the columns we need

                if hugo_col_num is None:

                    raise AttributeError("Unable to locate a column corresponding to \'%s\' in the input file \'%s\'" % (hugo_column, entrez_file))

                elif entrez_col_num is None:

                    raise AttributeError("Unable to locate a column corresponding to \'%s\' in the input file \'%s\'" % (entrez_column, entrez_file))

                # If we have made it this far, then the header was valid

                continue

            # Check the Approval Status column to see if the gene name has been withdrawn

            if status_col_num is not None:

                if cols[status_col_num] != "Approved":

                    continue

            # Parse the gene name and ID from the input file

            try:

                hugo_name = cols[hugo_col_num]

                entrez_id = cols[entrez_col_num]

                # Sanity check: The entrez ID is ALWAYS a number

                if not entrez_id.isdigit():

                    # If there is no entrez ID, skip this gene

                    if entrez_id == "":

                        skipped_genes.append(hugo_name)

                        continue

                    else:

                        raise TypeError("When parsing line %s of \'%s\', the Entrez gene ID \'%s\' does not appear to be a valid Entrez Gene ID" % (i, entrez_file, entrez_id))

            except IndexError as e:

                # If we end up in here, then there were too few columns in the current line to parse the IDs

                raise AttributeError("Unable to parse line %s of \'%s\': Line is truncated?" % (i, entrez_file)) from e

            # Sanity check: Make sure we haven't seen this gene name before

            if hugo_name in hugo_name_to_id:

                raise AttributeError("Gene \'%s\' appears duplicated in \'%s\'" % (hugo_name, entrez_file))

            hugo_name_to_id[hugo_name] = entrez_id

            # If the Previous symbols column was found, annotate these gene names with the same ID

            previous_names = cols[prev_name_col_num].split(",")

            if previous_names[0] == "":

                continue  # There are no previous names for this gene

            previous_names = [x.replace(" ", "") for x in previous_names]  # Remove extra spaces

            for hugo_name in previous_names:

                # If we have seen this gene name before, then don't overwrite it. Just assume that the first instance

                # encountered is correct. This likely isn't the case, but these cases should be so rare that they are

                # not worth worrying about

                if hugo_name in alt_hugo_name_to_id:

                    continue

                alt_hugo_name_to_id[hugo_name] = entrez_id

    # Final sanity check: As Hugo Symbols and Entrez IDs have a 1-1 mapping, make sure there are no duplicate Entrez IDs

    if len(hugo_name_to_id) != len(set(hugo_name_to_id.values())):

        num_ids = Counter(hugo_name_to_id.values())

        # For simplicity, just print out one Entrez ID which is duplicated

        bad_id = num_ids.most_common(1)[0][0]

        bad_count = num_ids.most_common(1)[0][1]

        raise AttributeError("Entrez gene ID \'\%s\' appears %s times in the input file" % (bad_id, bad_count))

    return hugo_name_to_id, alt_hugo_name_to_id

def load_subset_ids(id_file):

    """

    Loads a set of Entrez IDs from a specified file. One ID/line

    :param id_file:

    """

    subset_ids = []

    with open(id_file) as f:

        for line in f:

            line = line.rstrip("\n").rstrip("\r")

            if not line.isdigit():

                raise TypeError("Invalid Entrez ID when processing the --lyphmgen_genes file: %s" % line)

            subset_ids.append(line)

    subset_ids = set(subset_ids)

    return subset_ids

def generate_mut_flat(in_maf: str, seq_type: str, gene_ids: dict, out_mut_flat: str, out_gene_list: str, alt_gene_ids: dict = None, subset_ids: iter = None):

    """

    Converts a MAF file into the mutation flat file and gene list used by LymphGen

    WARNING: GRCh37/hg19 is currently only supported!

    An example of the output mutation FLAT file:

    Sample  ENTREZ.ID   Type    Location

    Sample1 604 MUTATION    187451395

    Sample1 604 MUTATION    187451408

    Sample1 613 MUTATION    23523596

    Sample1 970 TRUNC   6590925

    Sample1 1840    Synon   113495932

    An example of the output gene list:

    "ENTREZ.ID"

    60

    355

    567

    596

    604

    605

    613

    639

    673

    If targeted sequencing is specified, only the genes with at least one non-synonmymous mutation in the MAF file are used for

    the gene list. Otherwise, ALL genes are used.

    The following MAF columns are required: Hugo_Symbol, Variant_Classification, Tumor_Sample_Barcode. If NCBI_Build and

    Start_Position are found, the output mutation flat file will contain an additional column specifying "Location".

     If Tumor_Seq_Allele2 is provided, the MYD88 hotspot mutations will be annotated as "L265P"

    :param in_maf: A string containing a filepath to a MAF file containing the variants of interest

    :param seq_type: A string specifying the sequencing type used to identify mutations. Only "targeted", "exome", and "genome" are supported

    :param gene_ids: A dictionary containing the mapping between Hugo_Symbol: Entrez_Gene_ID.

    :param out_mut_flat: A string specifying the output path to the mutation flat file

    :param out_gene_list: A string specifying the output path to the gene list

    :param alt_gene_ids: An additional dictionary specifying alternate Hugo_Symbol: Entrez_Gene_ID mappings

    :param subset_ids: An interable specifying a list of Entrez IDs. Only mutations within these Entrez IDs will be output

    :return: A list specifying all samples analyzed for mutations

    """

    logging.info("Processing MAF file")

    # Non-synonmymous mutation types:

    non_syn = {"In_Frame_Del", "In_Frame_Ins", "Missense_Mutation"}

    trunc_mut = {"Frame_Shift_Del", "Frame_Shift_Ins", "Nonsense_Mutation", "Nonstop_Mutation", "Splice_Site", "Translation_Start_Site"}

    synon_mut = {"Silent", "5'UTR", "5'Flank", "Intron", "3'UTR"}  # Use for the "Synon" mutation type

    # Which samples are we analyzing?

    sample_list = SortedSet()  # Used to ensure the output files are in somewhat sorted order

    required_cols = ["Hugo_Symbol", "Variant_Classification", "Tumor_Sample_Barcode"]

    optional_cols = ["NCBI_Build", "Start_Position", "Tumor_Seq_Allele2"]

    header_cols = {}

    out_header_written = False

    out_header_cols = ["Sample", "ENTREZ.ID", "Type"]

    if alt_gene_ids is None:

        alt_gene_ids = {}

    # For targeted sequencing, which genes have we sequenced and seen mutations in?

    genes_seen = {}

    skipped_mut = []  # Store mutations in genes with no Entrez ID

    i = 0

    with open(in_maf) as f, open(out_mut_flat, "w") as o:

        for line in f:

            i += 1

            # Ignore comment lines

            if line[0] == "#":

                continue

            line = line.rstrip("\n").rstrip("\r")

            cols = line.split("\t")

            # Skip empty lines

            if not line:

                continue

            # If we haven't parsed the header yet, assume we are parsing the first line of the file, and that is the header

            if not header_cols:

                j = 0

                for col in cols:

                    if col in required_cols:

                        header_cols[col] = j

                    elif col in optional_cols:

                        header_cols[col] = j

                    j += 1

                # Check that we have all required columns

                for col in required_cols:

                    if col not in header_cols:

                        raise AttributeError("Unable to locate a column specifying \'%s\' in the MAF file header" % col)

                # See which optional columns we have

                if "Start_Position" in header_cols:

                    out_header_cols.append("Location")

                    logging.info("\'Start_Position\' column found in MAF file. Output mutation flat file will be annotated with \'Position\'")

                if "Tumor_Seq_Allele2" in header_cols:

                    logging.info(

                        "\'Tumor_Seq_Allele2\' column found in MAF file. MYD88 hotspot mutations will be annotated as \'L265P\'")

                continue

            # Process mutations

            mut_attributes = {x: cols[y] for x, y in header_cols.items()}

            # Check genome build, as only GRCh37 is supported

            if "NCBI_Build" in mut_attributes and "Start_Position" in mut_attributes and mut_attributes["NCBI_Build"] != "GRCh37":

                raise NotImplementedError("Only GRCh37 is currently supported as a reference genome")

            # Lets get the easiest stuff out of the way first. Specify the output sample name and gene ID

            sample_name = mut_attributes["Tumor_Sample_Barcode"]

            if sample_name == "" or sample_name == ".":

                raise AttributeError("No tumor_sample_barcode was found when processing line %s of the MAF file" % i)

            hugo_name = mut_attributes["Hugo_Symbol"]

            try:

                # Skip intergenic mutations

                if hugo_name == "Unknown":

                    continue

                entrez_id = gene_ids[hugo_name]

            except KeyError:

                # If this gene does not have a Entrez ID, are we prehaps using an older Hugo_Symbol?

                if hugo_name in alt_gene_ids:

                    entrez_id = alt_gene_ids[hugo_name]

                else:

                    # Skip this mutation if there is no Entrez ID

                    skipped_mut.append(line)

                    continue

            # Obtain position (if Start_Position is specified)

            if "Start_Position" in mut_attributes:

                position = mut_attributes["Start_Position"]

            else:

                position = None

            # Finally, specify the variant type

            if mut_attributes["Variant_Classification"] in trunc_mut:

                genes_seen[entrez_id] = hugo_name  # This gene has a non-synonmymous mutation, so lets assume we have sequenced it

                type = "TRUNC"

            elif mut_attributes["Variant_Classification"] in non_syn:

                genes_seen[entrez_id] = hugo_name  # This gene has a non-synonmymous mutation, so lets assume we have sequenced it

                type = "MUTATION"

            elif mut_attributes["Variant_Classification"] in synon_mut:

                genes_seen[entrez_id] = hugo_name

                type = "Synon"

            else:

                # Ignore intronic mutations, 3'UTRs etc

                continue

            # If this mutation is in MYD88, does it affect the hotspot?

            if hugo_name == "MYD88" and "Tumor_Seq_Allele2" in mut_attributes:

                if mut_attributes["Start_Position"] == "38182641" and mut_attributes["Tumor_Seq_Allele2"] == "C":

                    type = "L265P"

            if sample_name not in sample_list:

                sample_list.add(sample_name)

            # If this gene isn't in the subset list provide it, skip it

            if subset_ids is not None and entrez_id not in subset_ids:

                continue

            # Finally, write this variant

            # Write a header line, if we have not done so already

            if not out_header_written:

                out_header_written = True

                o.write("\t".join(out_header_cols))

                o.write(os.linesep)

            out_line = [sample_name, entrez_id, type, position + os.linesep]

            o.write("\t".join(out_line))

    # Check to see that we actually converted mutations

    if len(genes_seen) == 0:

        raise AttributeError("No mutations were successfully converted. Check that the --entrez_ids file has valid Hugo Symbols matching the --maf file")

    elif skipped_mut:

        logging.warning("%s mutations in the MAF file were not converted. These either don't have a valid Entrez ID, or they were not in the --lymphgen_gene file" % len(skipped_mut))

    logging.info("Finished processing MAF file")

    logging.info("Generating gene list")

    # Generate the gene list file

    # This file a single column with the Entrez ID, as that is all LymphGen currently supports

    with open(out_gene_list, "w") as o:

        # Write header

        o.write("\"ENTREZ.ID\"" + os.linesep)

        # Write out all genes we have seen a mutation in

        for entrez_id in genes_seen.keys():

            if subset_ids is not None and entrez_id not in subset_ids:

                continue  # Skip these gene, as it wasn't in the --lymphgen_gene list

            o.write(entrez_id)

            o.write(os.linesep)

        if seq_type == "exome" or seq_type == "genome":

            # We have sequenced (effectively) all genes in the human genome. Write out all remaining genes

            for entrez_id in gene_ids.values():

                if entrez_id in genes_seen:  # We have already written out this gene. Skip it

                    continue

                if subset_ids is not None and entrez_id not in subset_ids:

                    continue  # Skip these gene, as it wasn't in the --lymphgen_gene list

                o.write(entrez_id)

                o.write(os.linesep)

    return list(sample_list)

def load_gene_coords_bed(bed_file, gene_ids, alt_gene_ids=None):

    """

    Load in the genomic coordinates of genes in the human genome from the specified BED4+ file.

    The following columns are required (no header)

    chrom   start   end gene

    A gene can be split across multiple BED entries (ex. exonic coordinates). In this case, the start and end of the first and last exon,

    respectively, will be used as the gene start/end coordinates

    :param bed_file: A string containing a filepath to a BED4+ file listing the positions of genes

    :return: A dictionary storing {gene_name: Gene()}

    """

    if alt_gene_ids is None:

        alt_gene_ids = {}

    gene_coords = {}

    skipped_genes = []

    i = 0

    with open(bed_file) as f:

        for line in f:

            i += 1

            line = line.rstrip("\n").rstrip("\r")  # Remove line endings

            cols = line.split("\t")

            # Skip empty lines

            if not line:

                continue

            # Since this BED file could contain more than 4 columns (ex. a BED6 or BED12 file), manually unpack and inspect the first

            # four columns, since those are the columns we care about

            try:

                chrom = cols[0]

                start = cols[1]

                end = cols[2]

                gene = cols[3]

            except IndexError:

                # This BED entry was trucated

                raise AttributeError("Unable to parse line %s of %s as it contains less than four columns (chrom, start, end, gene)" % (i, bed_file))

            # Check that the start and end are actually genomic coordinates

            if not start.isdigit():

                raise TypeError("When parsing line %s of %s, the start position \'%s\' is not a valid genomic coordinate" % (i, bed_file, start))

            if not end.isdigit():

                raise TypeError("When parsing line %s of %s, the end position \'%s\' is not a valid genomic coordinate" % (i, bed_file, end))

            start = int(start)

            end = int(end)

            chrom = chrom.replace("chr", "")

            # Is the gene name the Hugo Symbol or the Entrez gene ID?

            # If its an integer, lets assume its the Entrez ID

            # If its not, assume it is the Hugo Symbol and convert it to the Entrez ID

            if not gene.isdigit():

                try:

                    entrez_id = gene_ids[gene]

                except KeyError:

                    # If we can't find this Hugo Symbol in the default mappings, check the alt gene IDs

                    if gene in alt_gene_ids:

                        entrez_id = alt_gene_ids[gene]

                    else:

                        # This gene doesn't have an Entrez ID. Skip it

                        skipped_genes.append(gene)

                        continue

            else:

                entrez_id = gene

            # Have we seen this chromosome before?

            if chrom not in gene_coords:

                gene_coords[chrom] = {}

            # Have we seen this gene before?

            if entrez_id in gene_coords[chrom]:

                # If so, then we need to update the start/end of this gene based on this new BED entry

                existing_gene_entry = gene_coords[chrom][entrez_id]

                # Sanity check

                if chrom != existing_gene_entry.chrom:

                    raise AttributeError("We found two entries for gene \'%s\'. One is found on chromosome \'%s\', while the other is found on \'%s\'"

                                         % (gene, chrom, existing_gene_entry.chrom))

                if start < existing_gene_entry.start:

                    existing_gene_entry.start = start

                if end > existing_gene_entry.end:

                    existing_gene_entry.end = end

            else:

                # If we haven't seen this gene before, create a new entry for it

                gene_coords[chrom][entrez_id] = Gene(chrom, start, end, entrez_id)

    # Now that we have processed all genes, make sure we actually found Entrez IDs for a handful of genes

    # If we haven't, it means the input file likely didn't contain Hugo Symbols

    if len(gene_coords) == 0:

        raise AttributeError("No Hugo Symbols from the --genes file (column 4) were found in the --entrez_ids file. An example gene is %s" % (skipped_genes[0]))

    return gene_coords

def load_chrom_arm(arm_file):

    """

    Loads in the genomic coordinates corresponding to the arm of each chromosome

    The input should be a tab-delimited file with the following columns

    chromosome  start   end arm

    :param arm_file: A string containing a filepath to a tab-delimited file containing genomic coordinates for chromosome arms

    :return: A dictionary storing the genomic range of each chromosome, and each individual arm

    """

    arm_chrom = {}

    required_cols = ["chromosome", "start", "end", "arm"]

    header_cols = {}

    i = 0

    with open(arm_file) as f:

        for line in f:

            i += 1

            line = line.rstrip("\n").rstrip("\r")  # Remove line endings

            cols = line.split("\t")

            # Skip empty lines

            if not line:

                continue

            # If we haven't parsed the header yet, assume this is the first line of the file (aka the header)

            if not header_cols:

                j = 0

                for col in cols:

                    if col in required_cols:

                        header_cols[col] = j

                    j += 1

                # Check to make sure all required columns are found

                for col in required_cols:

                    if col not in header_cols:

                        raise AttributeError("Unable to locate column %s in the chromosome arm positions file \'%s\'" % (col, arm_file))

                # If we get this far, the header is valid

                continue

            try:

                arm_attributes = {x: cols[y] for x, y in header_cols.items()}

            except IndexError:

                # We were unable to find a required column. This line is likely truncated

                raise AttributeError("Unable to parse line %s of the chromosome arm positions file %s: The line appears truncated" % (i, arm_file))

            # Have we seen this chromosome before? If not, lets create an object to store its genomic features

            chrom = arm_attributes["chromosome"]

            # Strip chr prefix

            chrom = chrom.replace("chr", "")

            if chrom not in arm_chrom:

                arm_chrom[chrom] = Chromosome(chrom)

            # Save the arm coordinates in the chromosome

            try:

                arm_chrom[chrom].add(int(arm_attributes["start"]), int(arm_attributes["end"]), arm_attributes["arm"])

            except ValueError as e:

                raise TypeError("Unable to process line %s of \'%s\': start and end must be integers") from e

    return arm_chrom

def get_overlap_genes(chrom: str, start: int, end: int, copy_num: int, gene_cords: dict):

    """

    Find the genes which overlap a given segment

    This is a very brute-force approach which will check all genes on the target chromosome for overlap. Pre-sorting and bisecting genomic

    regions would be significantly faster, but 1) You need to handle overlapping genes, and 2) I am assuming the performance penalty won't

    matter too much.

    Note that gains and losses are handled differently if a segment partially overlaps a gene. If the event is a loss, the gene is included.

    If the event is a gain, the gene is NOT included, as that copy is likely not functional

    :param chrom: A string coresponding to the contig name

    :param start: An int specifying the start of the segment

    :param end: An int specifying the end of the segment

    :param copy_num: An integer specifying the copy number of this segment

    :param gene_cords: A dictionary storing the positions of genes, in the format {chromosome: {gene1: attr, gene2: attr...}}

    :return:

    """

    # Handle chr-prefix shenanigans

    is_chr_prefixed = next(iter(gene_cords.keys())).startswith("chr")

    if is_chr_prefixed and not chrom.startswith("chr"):

        chrom = "chr" + chrom

    elif not is_chr_prefixed and chrom.startswith("chr"):

        chrom = chrom.replace("chr", "")

    try:

        genes_on_chrom = gene_cords[chrom]

    except KeyError:

        # No genes are on this contig. This could be a user error, or this a contig with no annotated genes

        return []

    # Go through each gene on this chromosome and see if the coordinates overlap with our regions

    olap_genes = []

    for entrez_id, gene_info in genes_on_chrom.items():

        if start < gene_info.end and end > gene_info.start:

            # Overlap found

            # Now for the tricky part

            # If the segment partially overlaps this gene, then we need to handle gains and losses differently

            # Deleting or gaining half of a gene will likely cause it to no longer function

            # Thus, partial deletions of genes could be drivers, while partial gains of genes are likely never drivers

            if copy_num < 2:

                olap_genes.append(entrez_id)

            elif copy_num > 2:

                if start > gene_info.start or end < gene_info.end:

                    continue  # Partially overlapping

                else:

                    olap_genes.append(entrez_id)

            else:

                raise NotImplementedError("Not calculating overlapping genes for copy-neutral segments")

    return olap_genes

def generate_cnv_files(cnv_segs, gene_regions_bed, arm_regions, gene_ids, out_cnv_gene, out_cnv_arm, sample_ids, alt_gene_ids=None, subset_ids=None, input_log2: bool= False, focal_cn_thresh:int = 30000000):

    """

    Characterize focal and arm-level copy number events, and summarize them in the respective output files.

    For focal events (i.e. events smaller than the specified threshold), identify all genes which overlap the even, and save them to the out_cnv_gene file

    All events are used to flag chromosomes or individual chromosomal arms which are gained or amplified.

    The cnv_segs file should have the following columns (extra columns are ignored):

    Tumor_Sample_Barcode    chromosome  start"  end CN

    Where CN is the absolute copy number

    The gene_regions_bed can have multiple entries for each gene (ex. exons). The maximum and minimum of the cumulative entries is used to define the

    gene boundaries

    The arm_regions file should contain the following columns:

    chromosome  start   end arm

    :param cnv_segs: A string containing a filepath to a tab-delimited file containing copy number events

    :param gene_regions_bed: A string specifying a filepath to a BED4+ file containing gene positions

    :param arm_regions: A string specifying a filepath to a tab-delimited file specifying the positions of chromosomal arms

    :param gene_ids: A dictionary mapping {Hugo_Symbol: Entrez_ID}

    :param out_cnv_gene: A string specifying the cnv_flat output file

    :param out_cnv_arm: A string specifying the cnv_arm output file

    :param sample_ids: A list containing samples which have one or more somatic mutations

    :param alt_gene_ids: An additional dictionary mapping {Hugo_Symbol: Entrez_IDs}. Used if previous Hugo_Symbols were assigned to a gene

    :param subset_ids: An iterable listing a set of Entrez IDs. Only CNVs overlapping these genes will be output

    :param input_log2: Does the input file contain log2 ratios instead of absolute CN?

    :param focal_cn_thresh: The maximum size of an event for it to be considered "focal", in bp

    :return: A list of samples which have CNV information

    """

    logging.info("Processing copy-number segments")

    if input_log2:

        logging.debug("Converting from log2 ratios")

    # First things first, lets load in the gene regions file and figure out where each gene is

    gene_coords = load_gene_coords_bed(gene_regions_bed, gene_ids, alt_gene_ids=alt_gene_ids)

    # Load in the chromosome arm regions

    arm_coods = load_chrom_arm(arm_regions)

    # Process copy number segments

    required_cols = ["Tumor_Sample_Barcode", "chromosome", "start", "end", "CN"]

    header_cols = {}

    sample_cnvs = {}

    i = 0

    with open(cnv_segs) as f:

        for line in f:

            i += 1

            line = line.rstrip("\n").rstrip("\r")  # Remove line endings

            cols = line.split("\t")

            # Skip empty lines

            if not line:

                continue

            # If we haven't parsed the header yet, assume this is the first line of the file (aka the header)

            if not header_cols:

                j = 0

                for col in cols:

                    if col in required_cols:

                        header_cols[col] = j

                    j += 1

                # Check to make sure all required columns are found

                for col in required_cols:

                    if col not in header_cols:

                        raise AttributeError("Unable to locate column \'%s\' in the CNV segments file \'%s\'" % (col, cnv_segs))

                # If we get this far, the header is valid

                continue

            # Process this CNV entry