SciGroveBeta
Genetics

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

Matteo Di Bernardo, Iain M. Cheeseman

Featured July 22, 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AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content.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 called Affinage acts like a super-smart biologist, reading thousands of science papers to figure out exactly how almost every human gene works, then writing clear, reliable summaries that are better and more up-to-date than old databases.

In depth
The paper introduces Affinage, a two-stage large language model (LLM) pipeline that systematically extracts and synthesizes mechanistic gene annotations from primary literature. It uses biologist-designed prompts to ensure only direct experimental evidence is considered, overcoming issues of LLM overclaiming and non-reproducibility. The output is stored as structured, reusable records for nearly all human protein-coding genes, significantly improving upon existing curated databases.

Key Takeaways

  • 1
    Affinage is a two-stage LLM pipeline that performs genome-scale mechanistic gene annotation from primary literature.
  • 2
    It uses biologist-designed prompts to extract only direct experimental evidence, preventing LLM overclaiming and ensuring high fidelity.
  • 3
    The pipeline generates structured, reusable records for 19,293 human protein-coding genes, outperforming UniProt and identifying uncharacterized genes.

Conceptual Flow

HIGH LEVEL
1
How Affinage Works

The system first finds all relevant papers, then an AI reads them to pull out only solid facts, and finally another AI puts those facts together into a clear story about each gene.

Gene Name
Find Papers
Paper List
2
Affinage's Impact on Gene Knowledge

This new system creates much better and more complete descriptions for almost all human genes compared to older methods, even finding many genes whose functions were previously unknown.

Old Gene Info
New Gene Info
Compare Quality
Better Gene Details
More Gene Details