Genomics professor hails new gene regulation model: methods section is a gold mine
anshulkundaje · x · 2026-11-12
Stanford computational genomics professor Anshul Kundaje praised a newly released gene regulation model as a major leap over its predecessor (AG1): intricate, non-trivial innovation across the architecture, and a methods section he calls "a gold mine of little nuggets" that everyone should read. He notes the model's zero-shot imputation to unseen cell states, improved distal enhancer–gene regulation, and better variant-effect prediction, and reports his own team's tests on GWAS variant and gene prioritization outperforming prior models.
- Improved distal enhancer–gene expression regulation → better cell-type-specific expression prediction → better variant effect prediction
- Still room for gains on splicing and post-transcriptional processes (decay, APA), possibly via SpliceAI2-style long-read RNA objectives
- Imputation power limited to interpolation over training conditions; out-of-distribution states will still fail
More from AGI Musings
- Clara Shih: junior workers face the worst job market in 37 years — clarashih · 2026-11-12
- RLHF Co-creator: LLMs Are Built for Assistance, Picking the Right Task Is Key to Automation — FinanceYF5 · 2026-11-12
- He gave up Harvard Law and an Anthropic offer to fight AI takeover — and got mocked — AaronBergman18 · 2026-11-12
- Vertex exec calls AI-cures-all claims "complete BS": 50 AI firms, zero working drugs — AlexTensor · 2026-11-12
- tinygrad slams EA: 'any means justify their ends' culture doomed to an SBF-style blowup — mark_k · 2026-11-12
- A painter asks: when an AI image moves you, where does the meaning live? — indianbull88 · 2026-11-12