HLA-SPREAD: a natural language processing based resource for curating HLA association from PubMed abstracts

Area of Science:

• Immunogenetics
• Bioinformatics
• Computational Biology

Background:

• The Human Leukocyte Antigen (HLA) system's complexity and nomenclature pose challenges for integrating disease association data.
• Over 146,000 PubMed studies and 28,320 HLA alleles exist, highlighting the need for structured information access.

Purpose of the Study:

• To develop an automated pipeline, HLA-SPREAD, for structured information retrieval on HLA associations with diseases, Adverse Drug Reactions (ADRs), and transplantation.
• To create a user-friendly interface for accessing curated HLA data from diverse global populations.

Main Methods:

• Utilized Natural Language Processing (NLP) to extract HLA information from approximately 28 million PubMed abstracts.
• Employed Python scripts for data mining, curation, filtering, and categorization into 24 hierarchical groups.
• Applied Named Entity Recognition (NER) and semantic analysis to infer HLA associations.

Main Results:

• HLA-SPREAD provides structured information on Single Nucleotide Polymorphisms (SNPs), populations, resources, ADRs, and diseases.
• Identified markers associated with allelic/haplotypic associations in autoimmune, cancer, viral, and skin diseases, as well as transplantation outcomes and hypersensitivity ADRs.
• The resource covers data from 109 countries and 40 ethnic groups, offering insights into HLA gene-disease patterns.

Conclusions:

• HLA-SPREAD is a novel resource for uncovering patterns in HLA gene-disease associations.
• Facilitates retrieval of clinically relevant biomarkers and susceptible/risk alleles for various diseases and ADRs.
• The tool is accessible at http://hla-spread.igib.res.in/ to aid researchers globally.