Alphabet Inc. Class C (GOOGL.US) DeepMind Releases AlphaGenome 2: Able to Read DNA Cells, Variant Identification Capability Nearly Doubles

Alphabet (GOOGL.US)'s artificial intelligence unit DeepMind on Wednesday released its latest AI model, AlphaGenome 2, designed to interpret DNA cells. "By moving from broad tissue averages down to individual cell types and cell states, and redesigning the model architecture, AlphaGenome 2 achieves a step-change in variant prediction performance," DeepMind wrote in a blog post. "Importantly, it lets researchers predict the effects of variants in disease-relevant cells that would be missed in broad tissue models, and also predict effects in cell states and disease states the model has not seen before. Compared with the previous state-of-the-art model, AlphaGenome 2 identifies nearly twice as many variants that alter gene expression within cells." DeepMind further stated that AlphaGenome 2 has been trained on 68 tissues and more than 100 million single-cell atlases. As a result, it can transform static cell atlases into a predictive engine for genetics. "Moving beyond the original model's reliance on whole-tissue averages, AlphaGenome 2 can resolve activity within specialized cellssuch as specific heart muscle cells or blood immune cellshelping to uncover key DNA variants that act only in specific cell types," DeepMind added. The new AI model improves on the previous version in several respects, including predicting gene regulation in specific cells, enhancing variant effect prediction, and pinpointing disease-related variants. The company cited a recent study showing that regulatory switches drive inflammatory bowel disease by overactivating the ETS2 gene in intestinal immune cells, and that AlphaGenome 2 can computationally reproduce this finding. AlphaGenome 2 is currently available for non-commercial research through the AlphaGenome API, and is open to commercial research use in private preview.