pubmed_parser

by titipata

titipata / pubmed_parser

:clipboard: A Python Parser for PubMed Open-Access XML Subset and MEDLINE XML Dataset

227 Stars 90 Forks Last release: 9 months ago (v0.2.3) MIT License 265 Commits 5 Releases

Available items

No Items, yet!

The developer of this repository has not created any items for sale yet. Need a bug fixed? Help with integration? A different license? Create a request here:

Pubmed Parser: A Python Parser for PubMed Open-Access XML Subset and MEDLINE XML Dataset

License DOI DOI Build Status

Pubmed Parser is a Python library for parsing the PubMed Open-Access (OA) subset , MEDLINE XML repositories, and Entrez Programming Utilities (E-utils). It uses the

lxml
library to parse this information into a Python dictionary which can be easily used for research, such as in text mining and natural language processing pipelines.

For available APIs and details about the dataset, please see our wiki page or documentation page for more details. Below, we list some of the core funtionalities and code examples.

Available Parsers

  • path
    provided to a function can be the path to a compressed or uncompressed XML file. We provide example files in the
    data
    folder.
  • for website parsing, you should scrape with pause. Please see the copyright notice because your IP can get blocked if you try to download in bulk.

Below, we list available parsers from

pubmed_parser
.

Parse PubMed OA XML information

We created a simple parser for the PubMed Open Access Subset where you can give an XML path or string to the function called

parse_pubmed_xml
which will return a dictionary with the following information:
  • full_title
    : article's title
  • abstract
    : abstract
  • journal
    : Journal name
  • pmid
    : PubMed ID
  • pmc
    : PubMed Central ID
  • doi
    : DOI of the article
  • publisher_id
    : publisher ID
  • author_list
    : list of authors with affiliation keys in the following format
 [['last_name_1', 'first_name_1', 'aff_key_1'],
  ['last_name_1', 'first_name_1', 'aff_key_2'],
  ['last_name_2', 'first_name_2', 'aff_key_1'], ...]
  • affiliation_list
    : list of affiliation keys and affiliation strings in the following format
 [['aff_key_1', 'affiliation_1'],
  ['aff_key_2', 'affiliation_2'], ...]
  • publication_year
    : publication year
  • subjects
    : list of subjects listed in the article separated by semicolon. Sometimes, it only contains the type of the article, such as a research article, review proceedings, etc.
import pubmed_parser as pp
dict_out = pp.parse_pubmed_xml(path)

Parse PubMed OA citation references

The function

parse_pubmed_references
will process a Pubmed Open Access XML file and return a list of the PMIDs it cites. Each dictionary has keys as follows
  • pmid
    : PubMed ID of the article
  • pmc
    : PubMed Central ID of the article
  • article_title
    : title of cited article
  • journal
    : journal name
  • journal_type
    : type of journal
  • pmid_cited
    : PubMed ID of article that article cites
  • doi_cited
    : DOI of article that article cites
  • year
    : Publication year as it appears in the reference (may include letter suffix, e.g.2007a)
dicts_out = pp.parse_pubmed_references(path) # return list of dictionary

Parse PubMed OA images and captions

The function

parse_pubmed_caption
can parse image captions from a given path to XML file. It will return reference index that you can refer back to actual images. The function will return list of dictionary which has following keys
  • pmid
    : PubMed ID
  • pmc
    : PubMed Central ID
  • fig_caption
    : string of caption
  • fig_id
    : reference id for figure (use to refer in XML article)
  • fig_label
    : label of the figure
  • graphic_ref
    : reference to image file name provided from Pubmed OA
dicts_out = pp.parse_pubmed_caption(path) # return list of dictionary

Parse PubMed OA Paragraph

For someone who might be interested in parsing the text surrounding a citation, the library also provides that functionality. You can use

parse_pubmed_paragraph
to parse text and reference PMIDs. This function will return a list of dictionaries, where each entry will have following keys:
  • pmid
    : PubMed ID
  • pmc
    : PubMed Central ID
  • text
    : full text of the paragraph
  • reference_ids
    : list of reference code within that paragraph.

This IDs can merge with output from

parse_pubmed_references
.
  • section
    : section of paragraph (e.g. Background, Discussion, Appendix, etc.)
dicts_out = pp.parse_pubmed_paragraph('data/6605965a.nxml', all_paragraph=False)

Parse PubMed OA Table [WIP]

You can use

parse_pubmed_table
to parse table from XML file. This function will return list of dictionaries where each has following keys.
  • pmid
    : PubMed ID
  • pmc
    : PubMed Central ID
  • caption
    : caption of the table
  • label
    : lable of the table
  • table_columns
    : list of column name
  • table_values
    : list of values inside the table
  • table_xml
    : raw xml text of the table (return if
    return_xml=True
    )
dicts_out = pp.parse_pubmed_table('data/medline16n0902.xml.gz', return_xml=False)

Parse MEDLINE XML

MEDLINE XML has a different XML format than PubMed Open Access. The structure of XML files can be found in MEDLINE/PubMed DTD here. You can use the function

parse_medline_xml
to parse that format. This function will return list of dictionaries, where each element contains:
  • pmid
    : PubMed ID
  • pmc
    : PubMed Central ID
  • doi
    : DOI
  • other_id
    : Other IDs found, each separated by
    ;
  • title
    : title of the article
  • abstract
    : abstract of the article
  • authors
    : authors, each separated by
    ;
  • mesh_terms
    : list of MeSH terms with corresponding MeSH ID, each separated by
    ;
    e.g.
    'D000161:Acoustic Stimulation; D000328:Adult; ...
  • publication_types
    : list of publication type list each separated by
    ;
    e.g.
    'D016428:Journal Article'
  • keywords
    : list of keywords, each separated by
    ;
  • chemical_list
    : list of chemical terms, each separated by
    ;
  • pubdate
    : Publication date. Defaults to year information only.
  • journal
    : journal of the given paper
  • medline_ta
    : this is abbreviation of the journal name
  • nlm_unique_id
    : NLM unique identification
  • issn_linking
    : ISSN linkage, typically use to link with Web of Science dataset
  • country
    : Country extracted from journal information field
  • reference
    : string of PMID each separated by
    ;
    or list of references made to the article
  • delete
    : boolean if
    False
    means paper got updated so you might have two

XMLs for the same paper. You can delete the record of deleted paper because it got updated.

dicts_out = pp.parse_medline_xml('data/medline16n0902.xml.gz',
                                 year_info_only=False,
                                 nlm_category=False,
                                 author_list=False,
                                 reference_list=False) # return list of dictionary

To extract month and day information from PubDate, set

year_info_only=True
. We also allow parsing structured abstract and we can control display of each section or label by changing
nlm_category
argument.

Parse MEDLINE Grant ID

Use

parse_medline_grant_id
in order to parse MEDLINE grant IDs from XML file. This will return a list of dictionaries, each containing
  • pmid
    : PubMed ID
  • grant_id
    : Grant ID
  • grant_acronym
    : Acronym of grant
  • country
    : Country where grant funding from
  • agency
    : Grant agency

If no Grant ID is found, it will return

None

Parse MEDLINE XML from eutils website

You can use PubMed parser to parse XML file from E-Utilities using

parse_xml_web
. For this function, you can provide a single
pmid
as an input and get a dictionary with following keys
  • title
    : title
  • abstract
    : abstract
  • journal
    : journal
  • affiliation
    : affiliation of first author
  • authors
    : string of authors, separated by
    ;
  • year
    : Publication year
  • keywords
    : keywords or MESH terms of the article
dict_out = pp.parse_xml_web(pmid, save_xml=False)

Parse MEDLINE XML citations from website

The function

parse_citation_web
allows you to get the citations to a given PubMed ID or PubMed Central ID. This will return a dictionary which contains the following keys
  • pmc
    : PubMed Central ID
  • pmid
    : PubMed ID
  • doi
    : DOI of the article
  • n_citations
    : number of citations for given articles
  • pmc_cited
    : list of PMCs that cite the given PMC
dict_out = pp.parse_citation_web(doc_id, id_type='PMC')

Parse Outgoing XML citations from website

The function

parse_outgoing_citation_web
allows you to get the articles a given article cites, given a PubMed ID or PubMed Central ID. This will return a dictionary which contains the following keys
  • n_citations
    : number of cited articles
  • doc_id
    : the document identifier given
  • id_type
    : the type of identifier given. Either
    'PMID'
    or
    'PMC'
  • pmid_cited
    : list of PMIDs cited by the article
dict_out = pp.parse_outgoing_citation_web(doc_id, id_type='PMID')

Identifiers should be passed as strings. PubMed Central ID's are default, and should be passed as strings without the

'PMC'
prefix. If no citations are found, or if no article is found matching
doc_id
in the indicated database, it will return
None
.

Install Package

Install directly from the repository

pip install git+git://github.com/titipata/pubmed_parser.git

or clone the repository and install using

pip
git clone https://github.com/titipata/pubmed_parser
pip install ./pubmed_parser

You can test your installation by running

pytest --cov=pubmed_parser tests/ --verbose

Example snippet to parse PubMed OA dataset

An example usage is shown as follows

import pubmed_parser as pp
path_xml = pp.list_xml_path('data') # list all xml paths under directory
pubmed_dict = pp.parse_pubmed_xml(path_xml[0]) # dictionary output
print(pubmed_dict)

{'abstract': u"Background Despite identical genotypes and ...", 'affiliation_list': [['I1': 'Department of Biological Sciences, ...'], ['I2': 'Biology Department, Queens College, and the Graduate Center ...']], 'author_list': [['Dennehy', 'John J', 'I1'], ['Dennehy', 'John J', 'I2'], ['Wang', 'Ing-Nang', 'I1']], 'full_title': u'Factors influencing lysis time stochasticity in bacteriophage \u03bb', 'journal': 'BMC Microbiology', 'pmc': '3166277', 'pmid': '21810267', 'publication_year': '2011', 'publisher_id': '1471-2180-11-174', 'subjects': 'Research Article'}

Example Usage with PySpark

This is a snippet to parse all PubMed Open Access subset using PySpark 2.1

import os
import pubmed_parser as pp
from pyspark.sql import Row

path_all = pp.list_xml_path('/path/to/xml/folder/') path_rdd = spark.sparkContext.parallelize(path_all, numSlices=10000) parse_results_rdd = path_rdd.map(lambda x: Row(file_name=os.path.basename(x), **pp.parse_pubmed_xml(x))) pubmed_oa_df = parse_results_rdd.toDF() # Spark dataframe pubmed_oa_df_sel = pubmed_oa_df[['full_title', 'abstract', 'doi', 'file_name', 'pmc', 'pmid', 'publication_year', 'publisher_id', 'journal', 'subjects']] # select columns pubmed_oa_df_sel.write.parquet('pubmed_oa.parquet', mode='overwrite') # write dataframe

See scripts folder for more information.

Core Members

and contributors

Dependencies

Citation

If you use Pubmed Parser, please cite it from JOSS as follows

Achakulvisut et al., (2020). Pubmed Parser: A Python Parser for PubMed Open-Access XML Subset and MEDLINE XML Dataset XML Dataset. Journal of Open Source Software, 5(46), 1979, https://doi.org/10.21105/joss.01979

or using BibTex

@article{Achakulvisut2020,
  doi = {10.21105/joss.01979},
  url = {https://doi.org/10.21105/joss.01979},
  year = {2020},
  publisher = {The Open Journal},
  volume = {5},
  number = {46},
  pages = {1979},
  author = {Titipat Achakulvisut and Daniel Acuna and Konrad Kording},
  title = {Pubmed Parser: A Python Parser for PubMed Open-Access XML Subset and MEDLINE XML Dataset XML Dataset},
  journal = {Journal of Open Source Software}
}

Contributions

We welcome contributions from anyone who would like to improve Pubmed Parser. You can create GitHub issues to discuss questions or issues relating to the repository. We suggest you to read our Contributing Guidelines before creating issues, reporting bugs, or making a contribution to the repository.

Acknowledgement

This package is developed in Konrad Kording's Lab at the University of Pennsylvania. We would like to thank reviewers and the editor from JOSS including

tleonardi
,
timClicks
, and
majensen
. They made our repository much better!

License

MIT License Copyright (c) 2015-2020 Titipat Achakulvisut, Daniel E. Acuna

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.