SciGroveBeta
Genetics

Affinage: genome-scale mechanistic gene annotation from the published literature

Matteo Di Bernardo, Iain M. Cheeseman

Featured July 10, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A new AI system reads all scientific papers about genes, carefully picking out only proven facts, then uses those facts to write a clear, detailed story about what each gene does, making gene information much better and easier to use.

In depth
The paper introduces Affinage, a two-stage large language model (LLM) pipeline that performs genome-scale mechanistic gene annotation. It first uses a biologist-designed reading pass to extract only direct experimental evidence from primary literature for each gene, then a synthesis pass reasons over these findings to create a structured, reusable annotation. This approach overcomes limitations of existing curated databases and naive LLM usage by ensuring faithfulness to evidence and reproducibility.

Key Takeaways

  • 1
    The authors developed Affinage, a two-stage LLM pipeline that systematically extracts and synthesizes mechanistic gene function from primary literature for 19,293 human protein-coding genes.
  • 2
    Affinage employs a biologist-designed reading pass to strictly filter for direct experimental evidence, preventing LLMs from overclaiming or fabricating information, and a synthesis pass to create structured, literature-grounded records.
  • 3
    The pipeline significantly outperforms curated databases like UniProt in coverage and detail, achieving a 99.1% win rate in head-to-head comparisons and identifying thousands of previously uncharacterized genes.

Conceptual Flow

HIGH LEVEL
1
Methodology: How Affinage Works

The system first finds all relevant papers, then an AI carefully reads them to pull out only proven facts, and finally, another AI uses those facts to write a clear story about the gene.

Gene Name
Find & Filter Papers
Relevant Papers
2
Results: Impact on Gene Knowledge

This new system creates much more detailed and up-to-date gene information than old methods, helping scientists understand genes better, especially those previously unknown.

Old Gene Info
Compare & Improve
New Gene Info