AlphaGenome 2
A genome prediction model released by Google DeepMind. It predicts the effects of DNA sequences and variants on specific human cell types and diseases, focusing on gene regulation and the functional consequences of non-coding variants.
Core Facts
Google DeepMind has launched AlphaGenome Atlas, an AI-powered atlas for predicting the impact of human genome mutations.
The goal of AlphaGenome 2 is to predict the impact of DNA sequences and their variants on specific human cell types and diseases.
Timeline (last 90 days)
AlphaGenome 2's release announcement appeared on Twitter.
AlphaGenome 2's announcement did not state whether industry users can obtain access and whether commercial use is restricted.
Google released AlphaGenome 2 on Wednesday, aimed at disease research.
AlphaGenome 2 is distributed via API rather than local weights.
AlphaGenome 2 is provided through the AlphaGenome API and is open for non-commercial research use.
The AlphaGenome 2 paper found that whether a variant lies in an open chromatin region predicts model performance better than its linear distance to a gene: variants in open regions can be predicted well even when far from genes, whereas variants in compacted regions are predicted poorly.
In a case involving a locus associated with inflammatory bowel disease, AlphaGenome 2 predicted that a variant regulates a distant gene, and a team at the Crick Institute experimentally validated this prediction.
Projects such as the Human Cell Atlas have accumulated large amounts of single-cell data, providing cell-type-specific supervision signals for training AlphaGenome 2.
AlphaGenome 2's differentiated advantage depends on three points: generalization to rare cell types, modeling quality of long-range regulatory interactions, and validation rates in wet-lab experiments.
AlphaGenome 2 is the successor to the previous-generation model AlphaGenome and is an upgraded version in the AlphaGenome series.
9 more timeline events
All Facts (19)
Google DeepMind has launched AlphaGenome Atlas, an AI-powered atlas for predicting the impact of human genome mutations.
65%Partially VerifiedThe goal of AlphaGenome 2 is to predict the impact of DNA sequences and their variants on specific human cell types and diseases.
75%UnverifiedAlphaGenome 2's announcement did not state whether industry users can obtain access and whether commercial use is restricted.
50%UnverifiedAlphaGenome 2 is distributed via API rather than local weights.
50%UnverifiedAlphaGenome 2 is provided through the AlphaGenome API and is open for non-commercial research use.
50%UnverifiedThe AlphaGenome 2 paper found that whether a variant lies in an open chromatin region predicts model performance better than its linear distance to a gene: variants in open regions can be predicted well even when far from genes, whereas variants in compacted regions are predicted poorly.
50%UnverifiedProjects such as the Human Cell Atlas have accumulated large amounts of single-cell data, providing cell-type-specific supervision signals for training AlphaGenome 2.
50%UnverifiedDeepMind says AlphaGenome 2 is expected to provide clues to disease causes and point to new therapeutic paths.
50%UnverifiedThe source material did not disclose AlphaGenome 2's specific performance metrics, the number of cell types covered, or benchmark results.
50%UnverifiedAlphaGenome 2's differentiated advantage depends on three points: generalization to rare cell types, modeling quality of long-range regulatory interactions, and validation rates in wet-lab experiments.
50%UnverifiedA more realistic positioning for AlphaGenome 2 is as an efficient assistant tool for researchers at the hypothesis-generation stage, rather than an endpoint that replaces experiments.
50%UnverifiedAlphaGenome 2 is the successor to the previous-generation model AlphaGenome and is an upgraded version in the AlphaGenome series.
50%UnverifiedAlphaGenome 2's design logic continues the approach of directly predicting molecular phenotypes from genomic sequences, but further refines it to the level of cell-type differences.
50%UnverifiedOfficially, only a one-sentence core function description of AlphaGenome 2 has been provided so far, with no disclosure of model architecture, training data scale, evaluation benchmarks or specific performance metrics.
50%UnverifiedVerifiable public details on the prediction accuracy and generalization capability of AlphaGenome 2 are currently still lacking.
50%UnverifiedThe core capability of AlphaGenome Atlas comes from DeepMind's AlphaGenome model, which can predict the effects of variants on molecular phenotypes such as gene expression regulation and splicing signals without requiring wet-lab experiments to be performed one by one.
50%UnverifiedAlphaGenome 2's release announcement appeared on Twitter.
50%UnverifiedGoogle released AlphaGenome 2 on Wednesday, aimed at disease research.
50%UnverifiedIn a case involving a locus associated with inflammatory bowel disease, AlphaGenome 2 predicted that a variant regulates a distant gene, and a team at the Crick Institute experimentally validated this prediction.
50%