23 Daily — According to theverge.com, Google DeepMind unveiled a new AI model named AlphaGenome 2 on November 11, 2026, designed to precisely analyze how genetic variations affect different cell types. This successor to last year's AlphaGenome model offers scientists a finer-grained understanding of gene variant effects at the cellular level, providing crucial insights into disease mechanisms and guiding new therapeutic development.
Capturing Cell Type and State-Specific Genetic Effects
Although nearly all human cells share the same DNA, different cell types switch genes on and off in unique patterns and these patterns shift depending on the cell’s state and function. Trained on over 100 million individual cell profiles across 68 tissues, AlphaGenome 2 predicts the impact of genetic variants on specific cell types and states, delivering much more detailed insights compared to broader models. Google reports it identifies nearly twice the number of gene expression-changing variants compared to its predecessor and can predict effects on novel cell types and states without retraining, revealing disease-relevant effects otherwise unnoticed.
Transforming Cell Atlases into Predictive Genomic Tools
AlphaGenome 2 effectively converts large static single-cell atlases into a dynamic predictive engine for human genetics. It also detects gaps in experimental data, helping researchers prioritize which cell states to study next. Adam Gayoso, a DeepMind research scientist involved in the project, noted the breakthrough lies in linking human genetics directly to cell biology, enabling scientists to drill down on how variants identified in population studies manifest at the cellular level and potentially relate to disease.
Editor's take: Entrepreneurial Opportunities and Validation Approaches
AlphaGenome 2 represents a powerful tool for biotech entrepreneurs focused on precision medicine and drug discovery, as the model provides granular insight into gene regulation in specific cell types, enabling better target identification. Startups can access the AlphaGenome API—currently available in a limited non-commercial preview—to integrate and validate the model using their own experimental data. Validation can involve comparing predicted variant effects on gene expression across known cell types to assess model accuracy and practical utility. Future enhancements including expanded datasets and cross-species training may further increase AlphaGenome 2’s capabilities, offering startups valuable assets for innovation in biomedical research.