In jaytimm/PubmedMTK: PubMed Mining Toolkit

PubmedMTK

An R package for querying the PubMed database & parsing retrieved records. Toolkit facilitates batch API requests & the creation of custom corpora for NLP.

Installation

You can download the development version from GitHub with:

library(tidyverse)

devtools::install_github("jaytimm/PubmedMTK")

Usage

PubMed search

The pmtk_search_pubmed() function is meant for record-matching searches typically performed using the PubMed online interface. The search_term parameter specifies the query term; the fields parameter can be used to specify which fields to query.

s0 <- PubmedMTK::pmtk_search_pubmed(search_term = 'medical marijuana', 
                                    fields = c('TIAB','MH'))

Sample output:

head(s0)

Multiple search terms

ps <- PubmedMTK::pmtk_search_pubmed(
  search_term = c('political ideology',
                  'marijuana legalization',
                  'political theory',
                  'medical marijuana'),
  fields = c('TIAB','MH'))

The pmtk_crosstab_query can be used to build a cross-tab of PubMed search results for multiple search terms.

ps0 <- PubmedMTK::pmtk_crosstab_query(x = ps) 

ps0 %>% knitr::kable()

Retrieve and parse abstract data

For quicker abstract retrieval, be sure to get an API key.

# Set NCBI API key
key <- '4f47f85a9cc03c4031b3dc274c2840b06108'
#rentrez::set_entrez_key(key)

## LIST of xml elements -- 
# https://www.nlm.nih.gov/bsd/licensee/elements_alphabetical.html

sen_df <- PubmedMTK::pmtk_get_records2(pmids = unique(s0$pmid), 
                                       with_annotations = T,
                                       cores = 5, 
                                       ncbi_key = key) 

Sample record from output:

sen_df <- data.table::rbindlist(sen_df)

n <- 10
list(pmid = sen_df$pmid[n],
     year = sen_df$year[n],
     journal = sen_df$journal[n],
     articletitle = strwrap(sen_df$articletitle[n], width = 60),
     abstract = strwrap(sen_df$abstract[n], width = 60)[1:10])

Annotations

Annotations are included as a list-column, and can be easily extracted:

annotations <- data.table::rbindlist(sen_df$annotations)

annotations %>%
  filter(!is.na(Form)) %>%
  slice(1:10) %>%
  knitr::kable()

Citation data

The pmtk_get_icites function can be used to obtain citation data per PMID using NIH's Open Citation Collection and iCite.

Hutchins BI, Baker KL, Davis MT, Diwersy MA, Haque E, Harriman RM, et al. (2019) The NIH Open Citation Collection: A public access, broad coverage resource. PLoS Biol 17(10): e3000385. https://doi.org/10.1371/journal.pbio.3000385

The iCite API returns a host of descriptive/derived citation details per record.

citations <- PubmedMTK::pmtk_get_icites(pmids = ps$pmid, 
                                        cores = 6,
                                        ncbi_key = key)

citations %>% select(-citation_net) %>%
  slice(4) %>%
  t() %>% data.frame() %>%
  knitr::kable()

Referenced and cited-by PMIDs are returned by the function as a column-list of network edges.

citations$citation_net[[4]]

Affiliations

The pmtk_get_affiliations function extracts author and author affiliation information from PubMed records.

afffs <- PubmedMTK::pmtk_get_affiliations(pmids = s0$pmid)

afffs %>%
  bind_rows() %>%
  slice(1:10) %>%
  knitr::kable()

jaytimm/PubmedMTK documentation built on Sept. 25, 2022, 10:49 p.m.