AI

Google DeepMind unveils AlphaGenome 2, an AI model that predicts how DNA changes affect different types of cells

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 11/11/2026, 21:47:25 UTC.

Briefings

Google DeepMind released AlphaGenome 2, an AI model trained on over 100 million cell profiles across 68 tissues. It significantly improved gene activity prediction accuracy compared to its predecessor, achieving nearly double the variant detection rate in immune cell benchmarks. The model enables predictions for unseen cell states without retraining, marking a major advancement in single-cell genomics.

Key points

• AlphaGenome 2 was trained on more than 100 million individual cell profiles across 68 different tissues.
• In a benchmark covering 28 immune cell types, AlphaGenome 2 detected nearly twice as many gene activity variants as the original AlphaGenome fine-tuned on single-cell data.

Context

This release follows previous AI models from Google DeepMind, including AlphaProof Code and Gemini 4 Carbon, demonstrating a consistent trend of expanding biological data scope and predictive capabilities.

Limitations

The evidence is limited to a blog post summary from a Reddit thread; specific performance metrics, model architecture details, or validation results are not provided in the source text.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 12/11/2026, 03:27:30 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeieflpj77dijesdj3rlvtnu7yt3medjfvcr3zh3zxhau4u4cjr2uzi