What happens when you change every single DNA letter in the human genome? Google DeepMind’s new AlphaGenome Atlas has a guess.
The data product, announced Sept. 8, contains precomputed predictions for the molecular effects of roughly 9 billion single-nucleotide variants, or every possible single-letter change in the human genome. Built on AlphaGenome, an artificial intelligence model for predicting the effects of genetic variants, the Atlas is meant to give researchers a genome-wide view rather than requiring them to query variants one at a time.
Accompanying the Atlas is the AlphaGenome Variant Impact (AVI) score, a new metric that combines predictions from both the AlphaGenome and AlphaMissense models into one number (ranging from 0 to approximately 55)
“Now, researchers can rapidly rank variants and interpret their molecular effects at the same time,” the DeepMind team wrote in a Sept. 8 blog post, adding that “Each AVI score is also linked to distinct biological features driving it, such as the aspects of gene regulation predicted by AlphaGenome or the protein impact score from AlphaMissense.” AVI scores work across both coding and non-coding regions.
In a self-published preprint, researchers from DeepMind as well as leading academic institutions from around the world, including the Broad Institute’s Heidi Rehm, Anne O’Donnell-Luria, and Fei Chen, provided a look at several applications of the Atlas.
For example, using data from the GREGoR (Genomics Research to Elucidate the Genetics of Rare diseases) Consortium, AVI scores helped prioritize previously overlooked variants and identified a variant in the gene DNM1 linked to epileptic encephalopathy. AlphaGenome’s underlying predictions showed the variant created an incorrect splice site that led to an abnormal protein extension, a finding reportedly confirmed by experimental screens.
AVI performed better than other variant impact metrics, including Combined Annotation Dependent Depletion (CADD), in benchmarking tests, the preprint authors also suggested.
Google DeepMind said the Atlas is available for academic research through a free website portal, with non-commercial access starting today and commercial availability planned on Google Cloud.
Other Genomics News
Ultima Genomics, Nvidia Partner on Pangenome-Aware WGS Analysis
Ultima Genomics is collaborating with Nvidia to enable pangenome-aware whole genome sequencing analysis practical for large-scale whole-genome sequencing (WGS) programs including population studies, biobanks, and precision medicine initiatives.
FDA Approves Longitudinal Use of Guardant360 CDx with AstraZeneca Breast Cancer Drug
For the first time, the FDA has approved a liquid biopsy companion diagnostic that could be run multiple times.
ArgenTag announced Sept. 8 that its single-cell RNA library kit for long-read sequencing has been added to Oxford Nanopore Technologies’ compatible products program. The kit partitions cells and generates full length, barcoded cDNAs without a dedicated instrument.
“ArgenTag’s single-cell transcriptomic solutions enable researchers to expand their biological understanding through the exploration of full-length transcripts, isoforms, splice variation, and fusion events at single-cell resolution,” Patrick Murphy, ONT’s market segment manager for transcriptomics, said in a statement.
OGT announced the expansion of its SureSeq Universal NGS Unique Dual Index range, enabling laboratories to multiplex up to 384 uniquely indexed samples in a single sequencing run.
Ion Genomics News
Look out for a new podcast episode this Friday with Gordon Freedman, founder of MitoWorld, an online hub for research into mitochondria and their roles in health and disease.
That’s all for this week, as I’m technically on vacation.