# Claim table — Google DeepMind AlphaGenome Atlas YouTube source check

| Claim | Evidence label | Notes |
| --- | --- | --- |
| Google DeepMind released AlphaGenome Atlas with predictions for 9 billion possible single-letter DNA changes. | Supported | DeepMind’s Sept. 8, 2026 blog says the Atlas contains predictions for 9 billion single-nucleotide variants and is free for academic research. |
| The Atlas is a 1-petabyte dataset, more than 30 times larger than AlphaFold Database. | Supported as company/reporter figure | DeepMind and Nature report the 1-petabyte/30-times-larger framing. |
| It can help rare-disease researchers prioritize variants. | Supported with caveat | DeepMind/Nature describe a DNM1 epileptic encephalopathy example and AVI score use; this is prioritization for research and lab follow-up, not a standalone diagnosis. |
| The underlying AlphaGenome model is peer-reviewed. | Supported | The AlphaGenome model paper appears in Nature, “Advancing regulatory variant effect prediction with AlphaGenome,” DOI 10.1038/s41586-025-10014-0. |
| AlphaGenome/Atlas is validated for clinical use or can diagnose a patient from their genome. | Not supported | DeepMind states AlphaGenome is not intended as medical advice and has not been validated or approved for clinical use. |
| This is immediately a cancer test or treatment. | Not supported | It may help cancer researchers interpret regulatory variants, but it is a research tool, not a screening test, diagnostic device, or therapy. |
| The video’s “massive breakthrough” framing. | Mostly fair, but promotional | The release is significant and independently covered, but the article should preserve limits: prediction is not causation, lab validation is still needed, and personal-genome use is not ready. |