Is your feature request related to a problem? Please describe.
Assessing the functional impact of non-coding variants remains challenging.
Describe the solution you'd like
Integrate AlphaGenome Variant Impact (AVI) scores and/or their PHRED-scaled values into seqr, with links to the corresponding variant pages in Google DeepMind’s AlphaGenome Atlas.
Describe alternatives you've considered
/
Additional context
Last week, Google DeepMind released the AlphaGenome Atlas (https://deepmind.google.com/science/alphagenome/atlas), a platform providing precomputed predictions for the effects of approximately 9 billion possible single-nucleotide variants in the human genome. Based on deep learning, it predicts how DNA sequence changes affect gene expression, splicing, chromatin accessibility, transcription factor binding and other regulatory processes. These predictions are particularly useful for prioritising non-coding and regulatory variants, whose functional consequences are often difficult to interpret. The AlphaGenome Variant Impact (AVI) score summarises variant impact, while its PHRED-scaled representation facilitates ranking and comparison. Adding these annotations to seqr, with options for filtering and sorting, would help users identify potentially relevant variants more efficiently alongside existing genetic and phenotypic evidence. Direct links to the Atlas could further support interpretation by allowing users to explore the underlying molecular predictions. These scores should support prioritisation rather than serve as standalone evidence of pathogenicity. Reference: Cheng et al. (2026), AlphaGenome Atlas: in silico mutagenesis of the entire human genome improves prioritization and interpretation of non-coding variants (preprint).
Is your feature request related to a problem? Please describe.
Assessing the functional impact of non-coding variants remains challenging.
Describe the solution you'd like
Integrate AlphaGenome Variant Impact (AVI) scores and/or their PHRED-scaled values into seqr, with links to the corresponding variant pages in Google DeepMind’s AlphaGenome Atlas.
Describe alternatives you've considered
/
Additional context
Last week, Google DeepMind released the AlphaGenome Atlas (https://deepmind.google.com/science/alphagenome/atlas), a platform providing precomputed predictions for the effects of approximately 9 billion possible single-nucleotide variants in the human genome. Based on deep learning, it predicts how DNA sequence changes affect gene expression, splicing, chromatin accessibility, transcription factor binding and other regulatory processes. These predictions are particularly useful for prioritising non-coding and regulatory variants, whose functional consequences are often difficult to interpret. The AlphaGenome Variant Impact (AVI) score summarises variant impact, while its PHRED-scaled representation facilitates ranking and comparison. Adding these annotations to seqr, with options for filtering and sorting, would help users identify potentially relevant variants more efficiently alongside existing genetic and phenotypic evidence. Direct links to the Atlas could further support interpretation by allowing users to explore the underlying molecular predictions. These scores should support prioritisation rather than serve as standalone evidence of pathogenicity. Reference: Cheng et al. (2026), AlphaGenome Atlas: in silico mutagenesis of the entire human genome improves prioritization and interpretation of non-coding variants (preprint).