Google presents an improved version of its AI to advance genome understanding

Google DeepMind has announced AlphaGenome 2, its state-of-the-art artificial intelligence (AI) model for predicting how DNA sequences and their variations affect specific human cell types, providing scientists with a “powerful tool” to discover the cellular factors that cause diseases.

The model is trained on more than 100 million individual cell profiles from 68 tissues, including data from the “Human Cell Atlas,” an international initiative that is mapping all cell types of the human body to transform the understanding of health and disease.

In 2025, Google launched AlphaGenome to predict how DNA variants affect molecular activity across a wide range of tissues -in January of this year, the company published the details of this AI in the journal Nature-.

He then expanded its reach with AlphaGenome Atlas, a platform containing predictions for the 9 billion possible single-letter changes in the human genome.

These tools help researchers prioritize and interpret the effects of variants, driving discoveries in rare diseases, complex traits, and fundamental biology, according to DeepMind.

“A substantial improvement” in the predictive performance of the variants

AlphaGenome 2 represents “a substantial improvement” in the predictive performance of variants.

It allows researchers to predict the effects of these in cells relevant to the disease that would be overlooked in general tissue models, as well as to make predictions about cellular and disease states that the model had not seen before.

The new version identifies almost twice as many variants that modify gene expression in cells, compared to the previous model, the company explains in a statement.

AlphaGenome 2, which can analyze new experiments without the need for retraining, is now available for non-commercial research via the AlphaGenome API, and in private preview for commercial research.

As scientists map more of the human body at the single-cell level, artificial intelligence models will continue to gain power.

“We look forward to seeing how researchers across the community use AlphaGenome 2 and its improved, cell-type-specific predictions to uncover how genetic variants act in different cells, and what this reveals about fundamental biology and disease,” the company notes.

AlphaGenome and AlphaGenome 2 are intended for research purposes only and have not been validated or approved for clinical or diagnostic use.