aboutsummaryrefslogtreecommitdiff
path: root/workflows/pull-data/genbank/genbank-fetch-ids.py
blob: 1962daa0fd7eae214c5a8a40ba7b6e57db319abb (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
#!/usr/bin/env python3
#
# Find all genbank IDs
#
#   genbank-fetch-ids.py --max 1000 --skip pubseq_ids.txt
#
# See also directory .guix-run and README.md

BATCH_SIZE=5000

import argparse
import json
import os
import requests
import sys
import xml.etree.ElementTree as ET
from datetime import date, datetime
from dateutil.parser import parse

parser = argparse.ArgumentParser()
parser.add_argument('--max', type=int, help='Max queries', required=False)
parser.add_argument('--skip', type=str, help='File with ids to skip, 1 id per line', required=False)
args = parser.parse_args()

from Bio import Entrez
Entrez.email = 'another_email@gmail.com' # FIXME

# min_acceptable_collection_date = datetime(2019, 12, 1)

today_date = date.today().strftime("%Y.%m.%d")

skip = set()
if args.skip:
    with open(args.skip) as f:
        content = f.readlines()
        for line in content:
            skip.add(line.strip())

print(f"Skip size is {len(skip)}",file=sys.stderr)

# Try to search several strings
TERMS = ['SARS-CoV-2', 'SARS-CoV2', 'SARS CoV2', 'SARSCoV2', 'txid2697049[Organism]']
# Remove mRNAs, ncRNAs, Proteins, and predicted models (more information here: https://en.wikipedia.org/wiki/RefSeq) starting with
PREFIX = ['NM', 'NR', 'NP', 'XM', 'XR', 'XP', 'WP']

ids = set()
for term in TERMS:
    num_read = BATCH_SIZE
    retstart = 0
    while num_read == BATCH_SIZE:
        record = Entrez.read(
            Entrez.esearch(db='nuccore', term=term, idtype='acc',
                           retstart=retstart, retmax=BATCH_SIZE)
        )
        idlist = record['IdList']
        new_ids = set(idlist)
        num_read = len(new_ids)
        print(num_read,":",idlist[0],file=sys.stderr)
        retstart += num_read
        new_ids.difference_update(skip) # remove skip ids
        new_ids = set([id for id in new_ids if id[:2] not in PREFIX])
        ids.update(new_ids)             # add to total set
        print(f"Term: {term} --> #{len(new_ids)} new IDs ---> Total unique IDs #{len(ids)})",file=sys.stderr)
        if args.max and len(ids) > args.max:
            print(f"Stopping past #{args.max} items",file=sys.stderr)
            break

for id in ids:
    print(id)