Researchers at Google DeepMind have unveiled a new AI model designed to understand how changes in our DNA affect different cells, giving scientists a more detailed view of a notoriously difficult problem that could hold important clues about the causes of disease – and pave the way for new treatments.
AlphaGenome 2 is the successor to DeepMind’s genome prediction AI model, AlphaGenome, launched last year. While almost all cells in the human body carry the same DNA, different types of cells – from heart muscle cells and skin cells to immune cells fighting an infection – turn different genes on and off, and these patterns can change depending on what a cell does. As a result, the same genetic variant can have very different effects depending on the type and state of the cell.
The new model is designed to capture these differences, predicting the effects of genetic variants on individual cell types and states. Google said in a blog post published Wednesday that it provides a much finer view than broader models like AlphaGenome, potentially revealing disease-related effects in specific cells that might otherwise go unnoticed.
AlphaGenome 2 was trained on over 100 million individual cell profiles across 68 tissues to give it this more detailed view. DeepMind says the model effectively turns static cell atlases – huge data repositories that map and catalog individual cells in detail – “into a predictive engine for human genetics”. According to the company, AlphaGenome 2 identifies nearly twice as many variants that could change gene expression in cells than its previous state-of-the-art model.
Notably, the researchers discovered that AlphaGenome 2 could also make predictions about new cell types and biological states without needing to be retrained. It could also identify gaps in experimental data, potentially helping scientists decide what to study next in the lab.
For Adam Gayoso, a researcher at Google DeepMind involved in the project, the big breakthrough consists of “linking human genetics to cell biology”. AlphaGenome 2 gives researchers a way to delve deeper into genetic variants identified in large populations, see what effects they may have at the cellular level, and potentially link those effects to disease, he said. The edge.
Žiga Avsec, head of genomics at Google DeepMind, said The edge that this specificity is important if you have a goal like understanding how to treat a particular disease. “When you develop drugs, you target specific proteins and specific cell types,” he explained, making this type of granularity vital.
Building a model like AlphaGenome 2 has only recently become possible thanks to new technologies that can capture individual cells and their states at high resolution and create the large data sets needed to train them, Avsec said. DeepMind is now working to improve its genetic models by generating more data, particularly on cellular states that are not well represented today. Avsec said future versions could also potentially be trained on data “from multiple species,” giving models a broader view of biology. Longer term, he said models like AlphaGenome 2 could aid efforts to create “virtual cells,” digital simulations designed to mimic biological systems.
Although Avsec has stated that it considers AlphaGenome 2 a successor to AlphaGenome, it will not replace it anytime soon. He said his more detailed forecasts have a computational cost and are “more expensive” to produce than those of his predecessor, he said. These requirements would have made it impossible to use AlphaGenome 2 to produce the genome-wide atlas covering “all possible DNA letter changes in the human genome” that DeepMind released in September.
Google says it is making AlphaGenome 2 available for non-commercial research via the AlphaGenome API starting today, with a limited “private preview” for commercial research.
The model is the latest in a growing set of efforts by Google and rival AI labs to use the technology to solve major problems in science and medicine. Pushmeet Kohli, who leads DeepMind’s AI for Scientific Research, hinted that more impressive scientific results were on the way and that the organization is expected to soon open an automated materials science lab in the UK. Anthropic, meanwhile, experimented with drug design and gave Claude a wet lab, while OpenAI eliminated an impressive amount of long-standing problems in math and related fields.
Gn tech
