Integrated Retrieval and Analysis of 'PubMed', 'NIH', and 'NLM' Literature Data.

puremoe

puremoe provides a single, PMID-centered interface to PubMed and the wider NIH/NLM data stack — NIH iCite, PubTator3, and PMC. A PubMed search resolves to a set of PMIDs, which one retrieval function uses to assemble data frames from each service: metadata and abstracts, citation data, entity annotations, or full text. A local analysis layer then operates on those tables with no further API calls: corpus expansion and citation networks from iCite links, sentence-anchored PubTator co-occurrence and relation evidence, and MeSH descriptor keyness against PubMed-wide frequencies.

Installation

From CRAN:

install.packages("puremoe")

Development version:

remotes::install_github("jaytimm/puremoe")

The puremoe API

Search

• search_pubmed(query, ...) -- PubMed query string → character vector of PMIDs. Accepts standard PubMed syntax: field tags ([TiAb], [MeSH Terms], [DP]), Boolean operators, wildcards.

Retrieve

get_records(pmids, endpoint, cores, sleep, ncbi_key) -- the single retrieval function. Pass PMIDs and name an endpoint; get back a data.table.

Analyze

Functions that transform already-retrieved tables -- no additional API calls.

• citation_snowball(icites, direction, min_links, max_nodes) -- expand a corpus one hop along iCite citation links; returns a ranked candidate table with audit columns (seed, cited_links, citing_links, link_count).
• citation_network(icites) -- convert an icites table into nodes + edges (within-corpus citations only), carrying RCR and clinical flags as node attributes; ready for igraph/tidygraph.
• pubtator_context(pubtator) -- add sentence IDs, sentence-relative entity spans, relation entity labels, relation sentence anchors, and a sentence lookup table to PubTator output.
• pubtator_cooccurrence(ctx, window, by) -- count entity pairs co-occurring within or across sentences in a pubtator_context() result.
• pubtator_network(ctx) -- convert PubTator relations into nodes, edges, and lean evidence tables for graph workflows and edge inspection.
• mesh_keyness(records, measure) -- score a corpus's MeSH descriptors against PubMed-wide frequencies (log-odds or Dunning G2) to surface over- and under-represented terms.

ID conversion

• pmid_to_pmc(pmids, ...) -- PMID → PMC ID + DOI via the NCBI ID Converter.
• pmid_to_ftp(pmids, ...) -- PMID → PMC ID + open-access PMC Cloud Service XML URL; pass URLs to get_records(endpoint = "pmc_fulltext").

MeSH reference data

• data_mesh_thesaurus() -- MeSH descriptor thesaurus + supplementary concept records; one row per term/synonym.
• data_mesh_trees() -- MeSH hierarchical tree structure; tree numbers encode the classification path.
• data_mesh_frequencies -- bundled PubMed-wide descriptor frequencies for MeSH keyness baselines.

Utilities

• endpoint_info(endpoint) -- column definitions, rate limits, and notes for each endpoint. Returns a list or JSON; useful for tool schemas in LLM applications.

Vignettes

• Getting started -- search_pubmed() + all get_records() endpoints end-to-end
• MeSH tables -- thesaurus lookup, tree navigation, and PubMed-wide descriptor frequencies
• Citation snowballing -- expand a seed corpus along citation links, audit why each paper was admitted, and quantify the expansion space against PubMed-wide MeSH keyness
• PubTator context and relation networks -- add sentence context to PubTator annotations, count entity co-occurrence, and inspect relation-network evidence

License

MIT © Jason Timm

Citation

citation("puremoe")

Issues

Report bugs or request features at https://github.com/jaytimm/puremoe/issues.