PubMed Literature Mining Module

Overview

The PubMed module retrieves abstracts from PubMed and identifies drug mentions through dictionary-based matching against a curated anticancer drug list derived from ChEMBL.

How It Works

1. Altered genes are extracted from the input MAF file
2. For each gene, PubMed is queried: "<cancer type> AND <gene> AND (therapy OR treatment OR inhibitor)"
3. Retrieved abstracts are parsed for drug name mentions
4. Results are aggregated as gene-drug mention counts

Key Output Fields

Field	Description
variant	Gene symbol (Hugo_Symbol)
drug	Matched drug name (uppercase)
mention_count	Number of abstract-level co-mentions

Drug Dictionary

The curated drug list (data/chembl_anticancer_drugs.txt) is derived from ChEMBL and includes drugs identified through:

• Drug indications
• Pharmacological mechanisms of action
• ATC classifications
• Approval status

Limitations

Warning

Abstract-level co-mentions do not establish therapeutic relevance. This module should be interpreted as evidence retrieval rather than causal inference. The current implementation does not perform relation extraction, directionality classification, or study-type filtering.

Usage

from pubmed_module import run_pubmed

pubmed_df = run_pubmed(
    maf_path="sample.maf",
    cancer_type="NSCLC",
    output_path="pubmed_output.tsv",
    pubmed_token="YOUR_PUBMED_TOKEN"
)