Google DeepMind has announced "AlphaGenome Atlas," a database that predicts the effects of approximately 9 billion single nucleotide variants in the human genome on biological processes such as gene expression and RNA splicing.
AlphaGenome Atlas is a dataset on a scale of approximately one petabyte, pre-analyzed using the AI model "AlphaGenome." This scale is more than 30 times larger than the AlphaFold Database, which contains protein structure predictions. Researchers can search for variant prediction results for free via a web browser without requiring programming knowledge, and analysis via API is also available.
As part of this announcement, the "AlphaGenome Variant Impact (AVI) score" has been introduced to represent the importance of a variant as a single numerical value. This score integrates predictions from AlphaGenome with those from "AlphaMissense," which evaluates variants that alter proteins. This enables the ranking of variants in both coding and non-coding regions using a common metric.
Google DeepMind stated that the tool is already being utilized in research, such as elucidating the causes of rare diseases. For example, it was reported that a research team at the Broad Institute used the AVI score to narrow down variants, identifying a mutation in the DNM1 gene associated with epileptic encephalopathy and experimentally confirming its functional impact.
Source: