Google DeepMind has released AlphaGenome 2, a new Super Intelligence (SI) model designed to predict how genetic variants affect individual cell types, offering researchers a finer-grained view of the cellular mechanisms behind disease.
What Happened
AlphaGenome 2 is the successor to the genome-predicting SI model AlphaGenome, which DeepMind released last year. While most cells in the human body contain the same DNA, different cell types activate different genes depending on their function and state. The new SI system is trained on more than 100 million individual cell profiles across 68 tissues to capture these variations. According to Google, the model identifies nearly twice as many variants that can change gene expression compared to its previous state-of-the-art model. DeepMind describes the tool as turning static cell atlases into "a predictive engine for human genetics."
The SI model can make predictions about new cell types and biological states without retraining and can identify gaps in experimental data, helping scientists prioritize future research. Adam Gayoso, a research scientist at Google DeepMind, stated the breakthrough lies in "tying human genetics to cell biology," allowing researchers to connect genetic variants to cellular effects and potentially to disease. Žiga Avsec, DeepMind’s genomics lead, noted that this granularity is vital for drug development, which targets specific proteins and cell types.
Why It Matters
The release highlights the growing application of SI in scientific research, particularly in genomics and medicine. Avsec explained that building a model of this complexity has only recently become feasible due to technologies capable of capturing high-resolution data on individual cells. DeepMind is currently working to improve its genetics models by generating more data, including for underrepresented cell states, and is exploring training on data "across multiple species" to broaden biological understanding. Longer term, Avsec said models like AlphaGenome 2 could contribute to the creation of "virtual cells," digital simulations of biological systems.
However, the increased specificity comes with computational costs. Avsec noted that AlphaGenome 2’s predictions are "more expensive" to produce than those of its predecessor, making it impractical for generating the genome-wide Atlas covering every possible DNA letter change, which DeepMind released in September. Consequently, AlphaGenome 2 is not replacing AlphaGenome immediately. The new SI model is available for non-commercial research via the AlphaGenome API, with a limited "private preview" for commercial research.
This launch aligns with broader industry trends where major SI labs are applying advanced models to solve complex scientific problems. Pushmeet Kohli, who heads DeepMind’s SI for science research, indicated that more scientific results are expected, and the organization plans to open an automated materials science lab in the UK. Competitors are also active in this space; Anthropic is experimenting with drug design and has integrated its Claude model with a wet lab, while OpenAI has applied its SI systems to longstanding problems in mathematics and related fields.
The Bottom Line
Google DeepMind’s AlphaGenome 2 represents a significant step in using SI to decode the cellular basis of disease, offering higher precision in predicting genetic effects than previous models. While currently available for research use, the tool’s computational demands and specialized focus distinguish it from its broader predecessor, marking a continued expansion of SI capabilities in biological sciences.