Advertisement
Telegram ads for your business · TCA Media
Monday, 16 November 2026
The main stories.
Advertisement
Telegram ads for your business · TCA Media
Technology

Google DeepMind unveils AI for predicting gene function across different cell types

The AlphaGenome 2 model predicts how DNA changes may affect gene activity in specific cell types, such as heart muscle cells or immune system cells.

13 Nov, 16:06Read in Russian

Google DeepMind introduced the model on November 11. The previous version generated predictions for multiple tissues and cell lines simultaneously. AlphaGenome 2 accounts for a specified cell type or state, enabling assessment of where the same genetic variant may alter gene activity, even if this effect is difficult to detect in tissue-wide data.

For training, developers used over 107 million profiles of individual human and mouse cells from 68 tissue types. According to DeepMind, the model can also make predictions for cell states not present in the training data, using experimental results provided by researchers.

Such predictions help narrow down genetic variants worth testing in the lab when studying diseases. The first version of AlphaGenome remains available for large-scale studies where tissue-level predictions are sufficient.

AlphaGenome 2 is accessible via API for non-commercial research. For commercial research projects, DeepMind currently provides access in a closed preview mode.

Advertisement
Telegram ads for your business · TCA Media