import re
import time
from urllib.request import urlopen
import requests
from lxml import etree, html
from .utils import stringify_children
def load_xml(pmid, sleep=None):
"""
Load XML file from given pmid from eutils site
return a dictionary for given pmid and xml string from the site
Parameters
----------
pmid: (int, str)
String of integer of a PMID
sleep: int
how much time we want to wait until requesting new xml
default: None
Return
------
tree: Element
An eutils XML of a given PMID
"""
link = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&retmode=xml&id={}".format(
pmid
)
page = requests.get(link)
tree = html.fromstring(page.content)
if sleep is not None:
time.sleep(sleep)
return tree
def parse_pubmed_web_tree(tree):
"""
Giving a tree Element from eutils, return parsed dictionary from the tree
Parameters
----------
tree: Element
An lxml Element parsed from eutil website
Return
------
dict_out: dict
A parsed output in dictionary format, dictionary keys includes
'title', 'abstract', 'journal', 'affliation' (string of affiliation with ';' separated),
'authors' (string with ';' separated),
'keywords' (keywords and MeSH terms from an XML -- if MeSH term it will be 'MeSH descriptor':'MeSH name')
'doi', 'pii', 'year', 'language', 'version_id', 'version_date'
"""
if len(tree.xpath("//articletitle")) != 0:
title = " ".join([title.text for title in tree.xpath("//articletitle")])
elif len(tree.xpath("//booktitle")) != 0:
title = " ".join([title.text for title in tree.xpath("//booktitle")])
else:
title = ""
abstract_tree = tree.xpath("//abstract/abstracttext")
abstract = " ".join([stringify_children(a).strip() for a in abstract_tree])
if len(tree.xpath("//article//title")) != 0:
journal = ";".join([t.text.strip() for t in tree.xpath("//article//title")])
else:
journal = ""
pubdate = tree.xpath('//pubmeddata//history//pubmedpubdate[@pubstatus="medline"]')
pubdatebook = tree.xpath(
'//pubmedbookdata//history//pubmedpubdate[@pubstatus="medline"]'
)
if len(pubdate) >= 1 and pubdate[0].find("year") is not None:
year = pubdate[0].find("year").text
elif len(pubdatebook) >= 1 and pubdatebook[0].find("year") is not None:
year = pubdatebook[0].find("year").text
else:
year = ""
affiliations = list()
if tree.xpath("//affiliationinfo/affiliation") is not None:
for affil in tree.xpath("//affiliationinfo/affiliation"):
affiliations.append(affil.text)
affiliations_text = "; ".join(affiliations)
authors_tree = tree.xpath("//authorlist/author")
authors = list()
if authors_tree is not None:
for a in authors_tree:
firstname = (
a.find("forename").text if a.find("forename") is not None else ""
)
lastname = a.find("lastname").text if a.find("forename") is not None else ""
fullname = (firstname + " " + lastname).strip()
if fullname == "":
fullname = (
a.find("collectivename").text
if a.find("collectivename") is not None
else ""
)
authors.append(fullname)
authors_text = "; ".join(authors)
else:
authors_text = ""
keywords = ""
keywords_mesh = tree.xpath("//meshheadinglist//meshheading")
keywords_book = tree.xpath("//keywordlist//keyword")
if len(keywords_mesh) > 0:
mesh_terms_list = []
for m in keywords_mesh:
keyword = (
m.find("descriptorname").attrib.get("ui", "")
+ ":"
+ m.find("descriptorname").text
)
mesh_terms_list.append(keyword)
keywords = ";".join(mesh_terms_list)
elif len(keywords_book) > 0:
keywords = ";".join([m.text or "" for m in keywords_book])
else:
keywords = ""
doi = tree.xpath('//elocationid[@eidtype="doi"]')
try:
doi = doi[0].text
except IndexError:
doi = None
pii = tree.xpath('//elocationid[@eidtype="pii"]')
try:
pii = pii[0].text
except IndexError:
pii = None
language = tree.xpath("//language")
try:
language = language[0].text
except IndexError:
language = None
medline_citation = tree.xpath('//medlinecitation')
try:
version_id = medline_citation[0].attrib.get('versionid')
version_date = medline_citation[0].attrib.get('versiondate')
except IndexError:
version_id, version_date = None, None
dict_out = {
"title": title,
"abstract": abstract,
"journal": journal,
"affiliation": affiliations_text,
"authors": authors_text,
"keywords": keywords,
"doi": doi,
"pii": pii,
"year": year,
"language": language,
"version_id": version_id,
"version_date": version_date,
}
return dict_out
[docs]
def parse_xml_web(pmid, sleep=None, save_xml=False):
"""
Give an input PMID, load and parse XML using PubMed eutils
Parameters
----------
pmid: str
A string of PMID which you want to parse from eutils
sleep: int
An integer of how long you want to wait after parsing one PMID from eutils
default: None
save_xml: bool
if it is True, save an XML output as a string in the key ``xml`` in an output dictionary.
It is good to check the information in
if it is False, we won't save a full XML to an output
default: False
Return
------
dict_out: dict
A dictionary contains information of parsed XML from a given PMID
Examples
--------
>>> pubmed_parser.parse_xml_web(11360989, sleep=1, save_xml=False)
{
'title': 'Molecular biology and evolution. Can genes explain biological complexity?',
'abstract': '',
'journal': 'Science (New York, N.Y.)',
'affiliation': 'Collegium Budapest (Institute for Advanced Study), 2 Szentháromság u., H-1014 Budapest, Hungary. szathmary@colbud.hu',
'authors': 'E Szathmáry; F Jordán; C Pál',
'keywords': 'D000818:Animals;D005075:Biological Evolution;...',
'doi': '10.1126/science.1060852',
'year': '2001',
'version_id': None,
'version_date': None,
'pmid': '11360989'
}
"""
tree = load_xml(pmid, sleep=sleep)
dict_out = parse_pubmed_web_tree(tree)
dict_out["pmid"] = str(pmid)
if save_xml:
dict_out["xml"] = etree.tostring(tree)
return dict_out
def extract_citations(tree):
"""
Extract number of citations from a given eutils XML tree.
Parameters
----------
tree: Element
An lxml Element parsed from eutil website
Return
------
n_citations: int
Number of citations that an article get until parsed date. If no citations found, return 0
"""
citations_text = tree.xpath('//form/h2[@class="head"]/text()')[0]
n_citations = re.sub("Is Cited by the Following ", "", citations_text).split(" ")[0]
try:
n_citations = int(n_citations)
except:
n_citations = 0
return n_citations
def extract_pmc(citation):
"""
Extract PMC from a given eutils XML tree.
Parameters
----------
tree: Element
An lxml Element parsed from eutil website
Return
------
pmc: str
PubMed Central ID (PMC) of an article
"""
pmc_text = [c for c in citation.split("/") if c != ""][-1]
pmc = re.sub("PMC", "", pmc_text)
return pmc
def convert_document_id(doc_id, id_type="PMC"):
"""
Convert a given document id to dictionary of other id.
Please see http://www.ncbi.nlm.nih.gov/pmc/tools/id-converter-api/ for more info
Parameters
----------
doc_id: (int, str)
A string or integer of document ID
id_type: str
A document ID type corresponding to an input ``doc_id``
default: 'PMC'
options: 'PMID', 'DOI', or 'OTHER'
Return
------
output_dict: dict
A dictionary contains possible mapping of a given document ID including 'pmc', 'pmid', and 'doi'.
If the document ID cannot be found, this will return empty string instead
Examples
--------
>>> pubmed_parser.pubmed_web_parser.convert_document_id(6933944, id_type='PMC')
{'pmc': 'PMC6933944', 'pmid': '31624211', 'doi': '10.1126/science.aax1562'}
"""
doc_id = str(doc_id)
if id_type == "PMC":
doc_id = "PMC{}".format(doc_id)
pmc = doc_id
convert_link = "http://www.ncbi.nlm.nih.gov/pmc/utils/idconv/v1.0/?tool=my_tool&email=my_email@example.com&ids={}".format(
doc_id
)
elif id_type in ["PMID", "DOI", "OTHER"]:
convert_link = "http://www.ncbi.nlm.nih.gov/pmc/utils/idconv/v1.0/?tool=my_tool&email=my_email@example.com&ids={}".format(
doc_id
)
else:
raise ValueError("Give id_type from PMC or PMID or DOI or OTHER")
convert_page = requests.get(convert_link)
convert_tree = html.fromstring(convert_page.content)
record = convert_tree.find("record").attrib
if "status" in record or "pmcid" not in record:
raise ValueError("Cannot convert given document id to PMC")
if id_type in ["PMID", "DOI", "OTHER"]:
if "pmcid" in record:
pmc = record["pmcid"]
else:
pmc = ""
pmid = record["pmid"] if "pmid" in record else ""
doi = record["doi"] if "doi" in record else ""
return {"pmc": pmc, "pmid": pmid, "doi": doi}
[docs]
def parse_citation_web(doc_id, id_type="PMC"):
"""
Parse citations from given document id
Parameters
----------
doc_id: (str, int)
document id
id_type: str
corresponding type of doc_id. This can be a choice from the following ['PMC', 'PMID', 'DOI', 'OTHER']
Return
------
dict_out: dict
output is a dictionary contains following keys
'pmc' (Pubmed Central ID), 'pmid' (Pubmed ID),
'doi' (DOI of an article), 'n_citations' (number of citations for given articles),
'pmc_cited' (list of PMCs that cite the given PMC)
Examples
--------
>>> pubmed_parser.parse_citation_web(6933944, id_type='PMC')
{
'n_citations': 0,
'pmid': '31624211',
'pmc': '6933944',
'doi': '10.1126/science.aax1562',
'pmc_cited': []
}
"""
assert id_type in ["PMC", "PMID", "DOI", "OTHER"]
doc_id_dict = convert_document_id(doc_id, id_type=id_type)
pmc = doc_id_dict["pmc"]
link = "http://www.ncbi.nlm.nih.gov/pmc/articles/{}/citedby/".format(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/{}/citedby/?page={}".format(
pmc, 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,
"pmid": doc_id_dict["pmid"],
"pmc": re.sub("PMC", "", doc_id_dict["pmc"]),
"doi": doc_id_dict["doi"],
"pmc_cited": pmc_cited_all,
}
return dict_out
[docs]
def parse_outgoing_citation_web(doc_id, id_type="PMC"):
"""
A function to load citations from NCBI eutils API for a given document
Example URL:
https://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pmc&linkname=pmc_refs_pubmed&id=221212
Parameters
----------
doc_id: str
The document ID
id_type: str
A type of provided document ID, can be either 'PMC' or 'PMID'
Return
------
dict_out: dict
a dictionary containing the following keys 'n_citations' (number of citations for that article),
'doc_id' (the document ID number), 'id_type' (the type of document ID provided (PMCID or PMID)),
'pmid_cited' (a list of papers cited by the document as PMIDs)
>>> pubmed_parser.parse_outgoing_citation_web(6933944, id_type='PMC')
{
'n_citations': 11,
'doc_id': '6933944',
'id_type': 'PMC',
'pmid_cited': ['30705152', ..., ]
}
"""
doc_id = str(doc_id)
if id_type == "PMC":
db = "pmc"
linkname = "pmc_refs_pubmed"
elif id_type == "PMID":
db = "pubmed"
linkname = "pubmed_pubmed_refs"
else:
raise ValueError("Unsupported id_type `{}`".format(id_type))
link = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom={}&linkname={}&id={}".format(
db, linkname, doc_id
)
parser = etree.XMLParser()
with urlopen(link) as f:
tree = etree.parse(f, parser)
pmid_cited_all = tree.xpath("/eLinkResult/LinkSet/LinkSetDb/Link/Id/text()")
n_citations = len(pmid_cited_all)
if not n_citations: # If there are no citations, likely a bad doc_id
return None
dict_out = {
"n_citations": n_citations,
"doc_id": doc_id,
"id_type": id_type,
"pmid_cited": pmid_cited_all,
}
return dict_out