AlphaGenome
Google DeepMind's model that predicts what a change in a DNA sequence does to gene activity, and the one petabyte atlas of those predictions.
AlphaGenome reads a stretch of DNA and predicts how the cell will use it: which genes switch on, how strongly, and how the RNA gets cut and joined. Comparing the prediction for an original sequence against an altered one tells a researcher what a single letter change is likely to do.
In September 2026 DeepMind released AlphaGenome Atlas, which precomputes those predictions for all nine billion possible single-letter variants in the human genome, about one petabyte of data. It ships with the AlphaGenome Variant Impact score, a single number per variant that folds in conservation and protein damage signals, so a geneticist can rank suspects instead of running the model one at a time.