PMID to PMC API from Medline cannot convert all provided PMID

[titipata]

The API here cannot convert all PMID input. I was trying to parse citations from given set of PMIDs but it only returns subset of PMIDs that I provided. One possibility is to host pair of PMIDs/PMCs somewhere on the cloud and provide similar API or source file that user can use to convert PMID to PMC.

[titipata]

I uploaded PMID-PMC pairs (size of 91 MB, not bad not bad) where we can download as follow:

wget https://s3-us-west-2.amazonaws.com/science-of-science-bucket/nih/pmid_pmc_pair.csv

For given file, you can convert PMID to PMC on your own. From here, we can modify parse_citation_web function to receive just PMC be as below.

def parse_citation_web(pmc):
    """
    Parse citations from given PMC 
    Parameters
    ----------
    pmc: str, PMC of the document e.g. 'PMC1217341'
    Returns
    -------
    dict_out: dict, contains following keys
        pmc: Pubmed Central ID
        n_citations: number of citations for given articles
        pmc_cited: list of PMCs that cite the given PMC
    """

    link = "http://www.ncbi.nlm.nih.gov/pmc/articles/%s/citedby/" % str(pmc)
    page = requests.get(link)
    tree = html.fromstring(page.content)
    n_citations = extract_citations(tree)
    n_pages = int(n_citations/30) + 1

    pmc_cited_all = list() # all PMC cited
    citations = tree.xpath('//div[@class="rprt"]/div[@class="title"]/a/@href')[1::]
    pmc_cited = list(map(extract_pmc, citations))
    pmc_cited_all.extend(pmc_cited)
    if n_pages >= 2:
        for i in range(2, n_pages+1):
            link = "http://www.ncbi.nlm.nih.gov/pmc/articles/%s/citedby/?page=%s" % (pmc, str(i))
            page = requests.get(link)
            tree = html.fromstring(page.content)
            citations = tree.xpath('//div[@class="rprt"]/div[@class="title"]/a/@href')[1::]
            pmc_cited = list(map(extract_pmc, citations))
            pmc_cited_all.extend(pmc_cited)
    pmc_cited_all = [p for p in pmc_cited_all if p is not pmc]
    dict_out = {'n_citations': n_citations,
                'pmc': pmc,
                'pmc_cited': pmc_cited_all}
    return dict_out

[titipata]

Also, we also want to add Copyright Notice for scraping function so that users don't scrape too much and get blocked https://www.ncbi.nlm.nih.gov/pmc/about/copyright/#copy-PMC

[nick-hahner]

What about ftp://ftp.ncbi.nlm.nih.gov/pub/pmc/oa_file_list.csv ?
Not every PMCID has a corresponding PMID though according to this list.

[titipata]

@nick-hahner, nice! It contains ~ 1.8M rows of PMID/ PMC pairs of Open Access Subset. I'm still thinking about how to update the list regularly by not hurting the repository. I mean, I could upload PMC-PMID pairs from MEDLINE somewhere as I mentioned. Do you have any preference or suggestions on how to make it available on the repository?

[nick-hahner]

Actually this file is probably better with 4,892,265 rows:
ftp://ftp.ncbi.nlm.nih.gov/pub/pmc/PMC-ids.csv.gz

First rsync or wget -c -N ... the file to some directory like ~/.pp_data
Then you can use an sqlite3 db

# Create an indexed sqlite db 
import pandas as pd
from sqlalchemy import create_engine
engine = create_engine('sqlite:///pmid_to_pmcid.db')  # but some better location
df = pd.read_csv('~/.pp_data/PMC-ids.csv.gz', dtype=str)
df[['PMCID', 'PMID']].to_sql('pmc_pmid', engine, index=False, if_exists='replace')
engine.execute('create index pmc_idx on pmc_pmid(PMCID)')
engine.execute('create index pmid_idx on pmc_pmid(PMID)')

# then later you can fetch like so:
from sqlalchemy import create_engine, text as sqa_text
def get_pmcid_from_pmid(pmid):
    engine = create_engine('sqlite:///pmid_to_pmcid.db')
    ret = engine.execute(sqa_text('select pmcid from pmc_pmid where pmid = :pmid;'), pmcid=pmcid).fetchone()
    return ret[0] if ret else None

How's that sound?

[chengkun-wu]

@nick-hahner Yes! I used the ftp://ftp.ncbi.nlm.nih.gov/pub/pmc/PMC-ids.csv.gz file for my local conversion.