AlphaGenome 2 Aims to Help Researchers Read the Effects of DNA Variation
Google DeepMind’s AlphaGenome 2 is designed to predict how DNA sequences and genetic variants influence activity across different human cell types, supporting research into disease biology.

Google DeepMind has introduced AlphaGenome 2, a DNA sequence model designed to help researchers understand how genetic variations influence the activity of different human cell types. The system aims to connect information in DNA sequences with the way cells regulate genes, a process that is central to understanding biology and many diseases.
Why cell-specific gene activity matters
Nearly every cell in the human body carries the same DNA, but cells do not use every gene in the same way. A heart muscle cell, a neuron and an immune cell have different jobs, and each activates or suppresses particular genes to perform them. Changes in the regulatory regions of DNA can influence when and where genes are switched on.
Understanding these patterns is difficult because the effects of a genetic variant can depend on cell type and biological context. Researchers need ways to predict how sequence changes may influence gene regulation before testing hypotheses in laboratory experiments.
What the model is designed to do
DeepMind describes AlphaGenome 2 as a next-generation model that predicts how DNA sequences and their variations affect specific human cell types. The intended use is to help scientists identify cellular mechanisms that may explain how genetic differences contribute to disease. Such predictions can help prioritise experiments and generate testable hypotheses.
AI models in genomics do not replace laboratory validation. A prediction is not proof that a variant causes a condition, and biological systems are more complex than any single computational model can fully represent. Researchers must compare predictions with experimental evidence and account for limitations in available datasets.
Potential value for biomedical research
Tools that help connect genetic variation with cell-specific activity could support studies of rare diseases, common conditions and basic cellular biology. They may also help researchers understand why the same genetic change can have different effects across tissues or individuals.
The important measure of AlphaGenome 2 will be how well its predictions generalise beyond the data used to build it and whether independent teams can use it to produce reliable findings. Its promise lies in helping researchers navigate the enormous complexity of the genome, while keeping experimental science and careful interpretation at the centre of discovery.
Sources
Related stories

Dust Grains From Earth’s Upper Atmosphere Point to Ancient Kuiper Belt Ice
NASA researchers analysed 12 tiny dust particles collected from Earth's stratosphere and found mineral and oxygen-isotope signatures consistent with ancient water in the Kuiper Belt. The findings offer clues to how material was distributed in the early solar system.

India and EU finish Horizon Europe talks, opening a path to direct research funding
India and the European Union have concluded negotiations on India's association with Horizon Europe, but researchers will gain access to the new funding arrangements only after the agreement is signed and procedures are completed.

Tiny crystals point to an ancient mountain system hidden beneath Antarctic ice
Researchers analysed 1,712 zircon grains from Antarctic sediments and found evidence of Himalayan-scale mountain building during Gondwana's formation, with possible links to ancient ocean chemistry.

India Unveils Indigenous H5N1 Vaccine for Poultry After Bhopal Trials
India has unveiled an inactivated H5N1 vaccine developed by ICAR-NIHSAD in Bhopal for chickens. The vaccine targets clade 2.3.2.1a, and institute-reported trials found protection lasting up to six months, although wider rollout arrangements have not yet been detailed.
