Published

DeepMind releases AlphaGenome 2, moving from tissues to individual cell types to predict how DNA variants drive disease

On 11 November Google DeepMind released AlphaGenome 2, which predicts how DNA sequences and their variants affect specific human cell types. DeepMind says it identifies nearly twice as many variants that change gene expression in cells as its previous model. It is available today for non-commercial research through the AlphaGenome API and in private preview for commercial research.

The first AlphaGenome took a million-letter DNA sequence and predicted many tissues in parallel, which made it hard to separate closely related cell types. AlphaGenome 2 is a conditional model: alongside the DNA sequence it takes the target tissue, cell type or cell state as input and predicts one condition at a time. It was trained on bulk and single-cell data from human and mouse, including more than 107 million individual cell profiles across 68 tissues, and adds DNA methylation to readouts such as gene expression, histone modifications and chromatin accessibility.

According to DeepMind, on the TenK10K single-cell dataset of 28 immune cell types, AlphaGenome 2 identified 30% of causal variants versus 16% for the original model fine-tuned on single-cell data and other specialised tools, and 43% versus 23% of directional gene effects. It also reproduced computationally a study finding that a distant regulatory variant drives inflammatory bowel disease by overactivating the ETS2 gene in gut macrophages. Researchers can upload single-cell data from just a few dozen cells and the model adapts to unseen cell states without retraining.

So researchers do not have to run large computations themselves, DeepMind precomputed predictions for trait-associated variants from genome-wide association studies across hundreds of human cell and tissue conditions, covering complex traits from height to type 2 diabetes. DeepMind says it is working with partners including the Wellcome Sanger Institute to cover more cell types.

Sources

  1. AlphaGenome 2: Deciphering how cells read DNA to better understand disease — Google DeepMind

Topics:Google

Related stories

Open this story in Shift →