Call genenames.org Biomart xml query service to get gene names, symbols and aliases for a given list of Ensembl gene ids

biomartxmlquery_gene_names_symbols.py

#!/usr/bin/python
"""
Script to call genenames.org Biomart xml query service to get gene names
and symbols for a given list of Ensembl gene ids

Usage:
 > ./biomartxmlquery_gene_names_symbols.py --inputFile ensgeneids.txt > results.tsv

Input file: one Ensembl gene id per line
"""

import argparse
import os, io, csv, sys
import httplib2 as http
if sys.version < '3':
    from urlparse import urlparse
    from urllib import urlencode
else:
    from urllib.parse import urlparse, urlencode


headers = {"Content-Type": "application/x-www-form-urlencoded"}
uri = 'http://biomart.genenames.org/martservice/results'
path = '/fetch/ensembl_gene_id/'
parser = argparse.ArgumentParser()
parser.add_argument('--inputFile', help='Input File', required=False,
                    default="ensemblgeneids.txt")
args = parser.parse_args()
genes = []
query = "<!DOCTYPE Query>\
<Query client=\"biomartclient\" processor=\"TSV\" limit=\"-1\" header=\"1\">\
<Dataset name=\"hgnc_gene_mart\" config=\"hgnc_gene_config\">\
<Filter name=\"hgnc_gene__ensembl_gene__ensembl_gene_id_104\"\
 value=\"comma_separated_ensembl_ids\" filter_list=\"\"/>\
 <Attribute name=\"hgnc_gene__ensembl_gene__ensembl_gene_id_104\"/>\
 <Attribute name=\"hgnc_gene__approved_symbol_1010\"/>\
 <Attribute name=\"hgnc_gene__approved_name_1010\"/>\
 <Attribute name=\"hgnc_gene__hgnc_alias_symbol__alias_symbol_108\"/>\
 <Attribute name=\"hgnc_gene__hgnc_alias_name__alias_name_107\"/>\
 </Dataset>\
 </Query>"


def query_biomart(xmlquery):
    target = urlparse(uri)
    method = 'POST'
    h = http.Http()
    q = {"query": xmlquery}
    r = None
    response, content = h.request(target.geturl(),
                                  method, body=urlencode(q), headers=headers)
    if response['status'] == '200':
        r = content.decode('utf-8')
    else:
        print('Biomart query has returned error: ' + response['status'])
        exit(-1)
    return r


# Customize this function if you have any additional text in input file
# other than Ensembl gene ids
def pruneline(line):
    if line.startswith("WinLen"):
        r = line[20:].replace("_MethylCounts.xls", "")
    else:
        r = line
    return r


def process_inputfile(inf):
    i = 0
    s = ""
    with open(inf) as infile:
        for line in infile:
            i += 1
            ensid = pruneline(line.strip())
            if i > 1: s += ','
            s += ensid
            genes.append(ensid)
    s= query.replace("comma_separated_ensembl_ids", s)
    return s


def process_biomart_results(br):
    i = 0
    previd = ""
    csvfile = io.StringIO(br)
    gmap = {}
    r = csv.reader(csvfile, delimiter='\t')
    for row in r:
        i += 1
        if i == 1: continue
        ensid = row[0]
        aps = row[1]
        apn = row[2]
        if previd != ensid:
            if i > 2:
                gmap[previd] = printline(l)
            l = (ensid, aps, apn, set(), set())

        if len(row) > 3 and len(row[3])>0:
            l[3].add(row[3])
        if len(row) > 4 and len(row[4])>0:
            l[4].add(row[4])
        previd=ensid
    gmap[previd] = printline(l)
    return gmap


def printline(l):
    s = str(l[0])
    s += "\t" + str(l[1])  # approved symbol
    s += "\t\"" + str(l[2]) + "\""  # approved name
    for r in l[3]:
        s += '\t'
        s += r

    for r in l[4]:
        s += '\t'
        s += '"' + r + '"'
    return s


def print_sorted_results(rmap):
    i = 1
    for gene in genes:
        if gene in rmap:
            print(str(i) + '\t' + rmap[gene])
        else:
            print(str(i) + '\t' + gene)
        i += 1
    return


xml_query = process_inputfile(args.inputFile)
br = query_biomart(xmlquery=xml_query)
rmap = process_biomart_results(br)
print("Line#\tEnsembl-gene-ID\tApproved-symbol\tApproved-name\tAlias-symbols"
      "\tAlias-names")
print_sorted_results(rmap)


# Not used by the current workflow
def get_biomart_data_curl(ensgid):
    cmd = "curl --data-urlencode query@query.xml http://biomart.genenames.org/martservice/results"
    os.system(cmd)