The Information Machine
FollowingDay 7in the edition since 12 Nov 20265 sourcesupdated 13 Nov 2026

Google DeepMind's AlphaGenome 2 release

The gist

Latest: DeepMind's AlphaGenome 2 Predicts Cell-Specific DNA Variant Effects

Cell-type-specific variant interpretation is a standing gap in genomics: a genetic association tells researchers a variant matters but not which cell type is involved. The model offers a computational approach to narrow that gap for research purposes, though it is not validated for clinical use.

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

The full picture

Google DeepMind released AlphaGenome 2 on November 11, 2026, a model that predicts how DNA sequence variations affect gene regulation in specific cell types. The model takes a DNA sequence and a target cell type or state as input and predicts regulatory effects for that condition. Trained on over 107 million individual cell profiles across 68 tissues, it identifies causal variants at 30% versus 16% for the prior AlphaGenome model on 28 immune cell types from the TenK10K dataset, and predicts effect direction at 43% versus 23%. Those benchmark numbers come from DeepMind's own evaluation, and the model is for research only, not validated for clinical or diagnostic use. DeepMind precomputed predictions for trait-linked variants across hundreds of conditions covering traits from height to type 2 diabetes. The model reproduced a published finding computationally, singling out a causal variant that selectively boosts the ETS2 gene in gut macrophages in inflammatory bowel disease, leaving other cell types and genes largely unaffected. Accepting single-cell RNA-seq data as conditioning input lets it generalize to entirely new, unseen cell types and disease states without retraining. DeepMind plans to integrate AlphaGenome 2 into its Antigravity agentic science platform alongside Gemini 4. The original AlphaGenome model has drawn independent use in neurodegenerative disease research, with Alzforum reporting that researchers used it to generate mechanistic hypotheses for variants from GWAS and rare-variant studies in Alzheimer's disease.

How it developed
12 November 2026in the 13 Nov edition

Reporting on DeepMind's November 11 release of AlphaGenome 2, a conditional genomic model predicting cell-type-specific effects of DNA variants, added that the model predicts effect direction at 43% versus 23% for the prior best on 28 immune cell types.

Coverage also noted DeepMind precomputed predictions for trait-linked variants across hundreds of conditions spanning height to type 2 diabetes, and that benchmarks come from DeepMind's own evaluation, with the model not validated for clinical or diagnostic use.

First citedThe Neuron
First in the edition · 12 November 2026
11 November 2026in the 12 Nov edition

Google DeepMind released AlphaGenome 2 on November 11, 2026, a conditional genomic model that takes both a DNA sequence and a target cell type as input and identifies causal variants in 30% of cases versus 16% for the prior model on the TenK10K immune cell dataset.

As a demonstration, DeepMind reported it reproduced a published finding that a variant drives inflammatory bowel disease by overactivating the ETS2 gene specifically in gut macrophages. Separately, researchers have applied the original AlphaGenome to Alzheimer's disease research: one study nominated candidate causal genes in a GWAS of argyrophilic grain disease, and another found an AD risk variant may alter transcription and splicing of the Fc receptor gene FCER1G. API outputs cannot be used to train other machine learning models, and commercial access is available via Google Cloud.

Sources

Get the daily email

When stories like this move, I write about them in an email each morning.

One free email each morning, around 8 AM US Eastern.

See the latest edition · Privacy