The human genome contains roughly 3 billion base pairs, making comprehensive interpretation far beyond what a person could read directly. Gareth Hawkes describes whole genome sequencing as a way to measure that full sequence and study how changes in genetic code affect observable human traits, while Samantha Bryen emphasizes how little is understood about the instructions distributed throughout it.
AlphaGenome analyzes a region of genome sequence and predicts the effects of a single mutation. The Atlas workflow condenses roughly 10,000 model outputs into an AVI variant impact score, with higher values indicating a greater likelihood of affecting phenotypes such as disease. The team says these scores have been precomputed for 9 billion possible single-letter changes in the human genome, helping researchers narrow the search for consequential variants.
AlphaGenome Atlas combines about a petabyte of prediction data with a genome browser intended to make the results accessible to biologists who do not code. The speakers frame the resource as an emerging blueprint for understanding how any base pair may influence human traits, and as one component of a broader agentic framework that could connect specialized systems such as AlphaFold and AlphaGenome for scientific research.
Watch on YouTube
