Google DeepMind has introduced AlphaGenome, an artificial intelligence (AI) creation that its developers believe could have a “transformative impact” on medical discovery by predicting how DNA mutations behave. The method also shows promise in finding genes associated to specific disorders and determining the etiology of rare diseases.
Since its inception in June, hundreds of researchers worldwide have used AlphaGenome to facilitate studies into “neurodegenerative diseases, infectious diseases, and cancer.” The model predicts how variations or mutations in DNA affect a variety of biological processes that regulate genes.
This ability could greatly aid researchers in improving genetic testing, identifying the exact origins of diseases, and developing novel treatments more quickly. Additionally, it should help researchers better grasp the human genome, which is the entire collection of DNA instructions present in every cell.
Both the mouse and human genomes were used to train AlphaGenome. The model can simultaneously anticipate 5,930 human or 1,128 mouse genetic signals, according to a report by the program’s creators published in the journal Nature. 25 out of 26 evaluations showed that these forecasts either equaled or outperformed the current state-of-the-art models.
Drug development is one of the many areas where AlphaGenome “could have a transformative impact,” according to Natasha Latysheva, a research engineer at Google DeepMind. “The idea here is that scientists could better pinpoint the genes and the cell types associated with the particular trait or disease by combining AlphaGenome predictions with large genetic association studies, like those from UK Biobank,” she stated. The discovery of drug targets and, eventually, the creation of new medications may benefit from this addition.
“Cancer patients may have multiple mutations at the same time, and it can be difficult to distinguish between the numerous passenger non-causal mutations and the causal driver mutations,” Ms. Latysheva added, elaborating on its potential in cancer research. AlphaGenome is a model that could assist in prioritizing lists of variations based on their likelihood of being functional, causative, and contributing to the illness.
AlphaGenome offers “interesting applications” in gene therapy, she continued, and may potentially be able to help determine the possible causes of rare disorders. “The concept here is that you can actually begin using a strong DNA sequence to function model to create completely new DNA sequences with certain desired qualities. For instance, you can attempt to create a sequence that only causes a particular gene to be activated in nerve cells and not in muscle cells.
Since June 2025, researchers have been able to interact with AlphaGenome through an application programming interface (API), which enables external software systems to use it. About 3,000 scientists from 160 countries have since performed one million API calls, according to Pushmeet Kohli, Google DeepMind’s vice president of science and strategic initiatives.
A commercial version of the AlphaGenome model and its rates is presently undergoing early testing, and the model and its rates are now available for non-commercial study. Researchers from prestigious universities, such as UCL, are already using the model “to advance research into areas including neurodegenerative diseases, infectious diseases and cancer,” according to Mr. Kohli.
Although “proteins are only one chapter of the biological story,” Mr. Kohli made an analogy while considering the larger scientific landscape, saying that “if proteins are the ingredients of life, then DNA is the recipe.”
“The Book of Life was provided by the Human Genome Project, but it was still challenging to read,” he continued. We still need to figure out the semantics, but we have the text. The next crucial area of study is comprehending the grammar of this genome, what is encoded in our DNA, and how it controls life.
The Wellcome Sanger Institute’s head of generative and synthetic genomics, Professor Ben Lehner, hailed AlphaGenome as “a great example of how AI is accelerating biological discovery and the development of therapeutics.” He emphasized: “A crucial first step in creating more effective treatments is figuring out the specific variations in our genes that determine our susceptibility to thousands of diseases. This will be made much simpler by AlphaGenome and similar models that aid in figuring out our genome’s regulatory code.
The Wellcome Sanger Institute examined AlphaGenome with 500,000 new trials, according to Professor Lehner, who confirmed that the results showed that it works “very well.” But he issued a warning: “AI models are only as good as the data that they are trained on.” Due to their small size and lack of standardization, the majority of biology data currently available is not highly appropriate for AI. The creation of the data needed to train the upcoming generation of even more potent AI models is currently the most significant problem, he said. “We must complete this task quickly, cost effectively, and in a way that makes the data and the models that are produced publicly accessible.”
The Francis Crick Institute’s head of genomics, Dr. Robert Goldstone, praised AlphaGenome as “a major milestone in the field of genomic AI.” According to him, this degree of resolution—especially for non-coding DNA—is a significant advancement that takes the technology from theoretical curiosity to real-world application, enabling researchers to model and explore the genetic causes of complex diseases programmatically. Although AlphaGenome is not a panacea for every biological problem, Dr. Goldstone came to the conclusion that it is a fundamental, excellent instrument that transforms the genome’s static code into a language that can be understood for studying.
