- AlphaGenome 2 is an upgraded DNA sequence model that predicts the impact of genetic variations on human cells.
- The model analyzes individual cell types, helping researchers examine genetic variations associated with diseases.
- In benchmark tests, AlphaGenome 2 identified more causal variations, enhancing the ability to assess the impact of gene activity.
- The model is available for non-commercial research use and has not yet received validation or approval for clinical applications.
Google DeepMind has launched AlphaGenome 2, an upgraded DNA sequence model designed to predict how genetic variations affect different types of human cells, aiding researchers in exploring the genetic mechanisms behind diseases. This model builds on the original AlphaGenome, which was capable of mapping molecular activity across a wide range of tissues, covering both coding and non-coding regions of the genome, thus providing a unified approach to assess the potential impacts of genetic variations. Earlier this year, the company also released AlphaGenome Atlas, an independent predictive resource that maps the effects of every possible single-nucleotide DNA change in the human genome.
Features and Capabilities of AlphaGenome 2
According to company information, AlphaGenome 2 is not limited to broad tissue-level averages but analyzes individual cell types and their states. This allows researchers to examine how genetic variations affect disease-related cells and the differences that may exist under various cellular conditions. The model is trained on over 17 million individual cell data points from human and mouse samples, covering 68 tissue types. Unlike its predecessor, the model takes specific biological conditions and DNA sequences into account during input.
Researchers can select a cell type from a reference catalog or use their own single-cell data to explore previously unseen cellular states. The model can also estimate missing experimental measurement data, such as chromatin accessibility, from other available data.
In benchmark tests using TenK10K, a single-cell dataset covering 28 major immune cell types, AlphaGenome 2 identified 30% of causal variations, compared to only 16% identified by the original model after adjustments on single-cell data and other specialized tools. The model also improved the ability to determine the direction of the impact of variations on gene activity. In 43% of cases, the model correctly identified whether a variation increased or decreased gene expression, while the original model and comparison tools achieved only 23% accuracy.
These results suggest that the updated system may help researchers narrow down the most likely genetic changes affecting disease-related processes in specific cell types.
Examples of Cell-Specific Predictions
Google DeepMind also demonstrated the model's ability to examine genetic effects under previously unencountered cellular conditions. In one example, researchers used single-cell data from individuals who received the COVID-19 mRNA vaccine. AlphaGenome 2 was able to distinguish activated immune T cells from resting T cells and found that a variation associated with severe COVID-19 had a greater impact on gene activity in activated cells. The company also computationally reproduced findings related to inflammatory bowel disease (IBD) using the model.
Variations located in distant regulatory regions of DNA were shown to promote the activity of the ETS2 gene in intestinal macrophages. The model predicts that this effect is concentrated in these cells rather than occurring broadly across other cell types. These examples illustrate how cell-specific predictions can help researchers link genetic variations to the cells and biological processes involved in diseases. However, computational predictions still require experimental validation.
AlphaGenome 2 is now available for non-commercial research use through the AlphaGenome API, while commercial researchers can access the model through a private preview. Google DeepMind states that the model is intended for research purposes and has not yet received validation or approval for clinical or diagnostic use.
The Importance of AlphaGenome 2 in Genetic Research
The launch of AlphaGenome 2 marks a significant advancement in genomic research, as it not only predicts the impact of genetic variations but also analyzes them for specific cell types. This cell-specific prediction helps researchers gain a more accurate understanding of the genetic basis of diseases and may facilitate the development of new therapies. Although the model shows potential for research purposes, further experimental validation is needed to confirm the accuracy and reliability of its predictions.

