Google DeepMind has unveiled a new artificial intelligence model designed to transform how scientists understand the human genome. The deep-learning system, called AlphaGenome, aims to uncover the biological role of non-coding DNA. It could help accelerate the development of treatments for genetic diseases.
External researchers have described the tool as a major scientific breakthrough. They say it allows scientists to study and simulate the genetic foundations of conditions. Many of these remain difficult to treat using conventional methods.
Speaking to journalists, Pushmeet Kohli, vice president of research at Google DeepMind, said that the first complete human genome map, published in 2003, provided scientists with the full DNA sequence. However, it did not give its deeper meaning.
He explained that while researchers have long known the three billion DNA letter pairs—A, T, C, and G—true progress depends on understanding how those letters function together. Kohli described this challenge as learning the “grammar” of the genome rather than simply reading its text.
His remarks accompanied a new peer-reviewed study published in Nature, where AlphaGenome’s design and capabilities were formally introduced.
Why Non-Coding DNA Matters
Only about two per cent of human DNA contains instructions for building proteins, which perform most essential biological functions. Scientists once dismissed the remaining 98 percent as “junk DNA” due to limited understanding.
Modern research now shows that this non-coding DNA plays a critical regulatory role. It helps control when genes activate, how strongly they express, and how cells behave across different tissues. Many disease-linked genetic variants also exist within these non-coding regions.
Our breakthrough AI model AlphaGenome is helping scientists understand our DNA, predict the molecular impact of genetic changes, and drive new biological discoveries. 🧬
Find out more in @Nature ↓ https://t.co/jvBLRXYzdj pic.twitter.com/WEL4Ptdv06
— Google DeepMind (@GoogleDeepMind) January 28, 2026AlphaGenome focuses specifically on decoding these regulatory sequences and identifying how they influence biological processes inside cells.
How AlphaGenome Works
The AlphaGenome model was trained on large public datasets that measure non-coding DNA activity across hundreds of human and mouse cell types. It can analyze DNA sequences up to one million letters long. Importantly, it does so while maintaining high predictive resolution.
According to lead author Žiga Avsec, long DNA sequences are essential for understanding the complete regulatory environment surrounding a single gene. The model predicts where genes start and stop. In addition, it predicts how RNA is produced and how small genetic changes affect cellular behavior.
This combination of length and detail sets AlphaGenome apart from earlier genome models. Those models typically sacrificed one capability for the other.
The AlphaGenome API is now powering over 1 million API calls per day from over 3000 total users across 160 countries.
Researchers are already using it to tackle some of the toughest challenges in biology. Here's how the model works ↓ pic.twitter.com/iYPr6aZjyg
— Google DeepMind (@GoogleDeepMind) January 28, 2026Part of Google’s Broader AI Science Effort
AlphaGenome builds on Google’s expanding portfolio of AI-driven scientific tools, including AlphaFold, which won the 2024 Nobel Prize in Chemistry for its impact on protein research.
Study co-author Natasha Latysheva said AlphaGenome can help scientists map functional DNA elements. Furthermore, it helps them understand their molecular roles with greater precision.
Google confirmed that more than 3,000 researchers across 160 countries have already tested the model. The company has made AlphaGenome freely available for non-commercial research and encouraged scientists to enhance it with additional data.
Read: Gmail Spam Filters Malfunction as Google Acknowledges Inbox Issues
Independent researchers have welcomed the tool while urging realistic expectations. Ben Lehner of the University of Cambridge said AlphaGenome performs strongly and could help identify genetic differences that influence disease risk.
However, he cautioned that AI systems remain limited by the quality of their training data. In his view, current genomic datasets still lack the depth needed for perfect predictions.
Robert Goldstone of the Francis Crick Institute echoed that view, noting that gene expression is also shaped by environmental factors beyond the model’s scope. Despite these limits, he described AlphaGenome as a breakthrough tool for studying the genetic roots of complex diseases.
