CryptoSlate
Google DeepMind launched AlphaGenome 2 on Nov. 11, 2026, a DNA prediction model that shows how genetic changes behave in specific cell types and states rather than producing fixed outputs across whole tissues. On TenK10K, a benchmark covering 28 primary immune-cell types, DeepMind reports the model identified 30% of causal variants, up from 16% for the original AlphaGenome fine-tuned on single-cell data.
The model was trained on bulk and single-cell data from human and mouse samples, including more than 107 million individual cell profiles. Researchers can also supply single-cell gene-activity data to condition predictions on newly observed states without retraining, letting the system predict properties that were never measured and prioritize which experiments to run.
Why it matters
The core upgrade is context. A tissue-level result can suggest a disease link while leaving the responsible cell unresolved; AlphaGenome 2 asks how the same stretch of DNA behaves in different cellular settings, surfacing effects confined to particular cells that broader tissue models miss.
DeepMind demonstrated the capability by computationally reproducing an established bowel-disease mechanism: a regulatory variant that raises ETS2 gene activity specifically in gut macrophages. The biology was first reported by Imperial College London, the Francis Crick Institute and UCL in June 2024, so the model's contribution is confirming the variant-gene-cell combination computationally rather than discovering it.
Market impact
AlphaGenome 2 is available through DeepMind's API for non-commercial research, with commercial access in private preview. Both AlphaGenome models are explicitly not validated or approved for clinical or diagnostic use, so the near-term opportunity is sharper experiment design: connecting a disease-associated signal to a specific variant, target gene and cell type before any lab work begins.
For genomics and drug-discovery pipelines, the reported 30% versus 16% jump is a developer-reported benchmark on variant identification, not a measure of diagnostic accuracy or personal disease risk.
Frequently asked questions
-
What is Google DeepMind's AlphaGenome 2?
AlphaGenome 2 is a DNA prediction model announced on Nov. 11, 2026, that predicts how genetic changes act in specific cell types and states, rather than producing fixed outputs across whole tissues like the original AlphaGenome.
-
How much better is AlphaGenome 2 than the original model?
On the TenK10K benchmark covering 28 primary immune-cell types, DeepMind reports AlphaGenome 2 identified 30% of causal variants, compared with 16% for the original AlphaGenome fine-tuned on single-cell data.
-
Can AlphaGenome 2 be used for medical diagnosis?
No. DeepMind states both AlphaGenome models are intended for research only and have not been validated or approved for clinical or diagnostic purposes, so the benchmark results measure variant identification, not disease risk.
-
What did AlphaGenome 2 reproduce in the bowel-disease example?
The model computationally reproduced an established finding that a regulatory variant increases ETS2 gene activity in gut macrophages. The mechanism was first reported by Imperial College London, the Francis Crick Institute and UCL in June 2024.
-
How can researchers access AlphaGenome 2?
The model is available through DeepMind's API for non-commercial research, with commercial research access in private preview as of the Nov. 11 announcement.
WatcherGuru
CoinTelegraph
CoinDesk
WuBlockchain
TheBlock