NAIROBI, Kenya- Google has unveiled a new artificial intelligence tool designed to help scientists better understand the human genome, a development experts say could accelerate research into genetic diseases and future treatments.
The tool, known as AlphaGenome, was introduced on Wednesday by Google DeepMind researchers and detailed in a study published in the journal Nature. It uses deep learning to analyse long DNA sequences and predict how genetic variations influence biological processes in the body.
According to Google, the technology aims to bridge the gap between having a complete genetic map and fully understanding how genes function in health and disease.
How AlphaGenome Works
Speaking to journalists, Google DeepMind Vice President of Research Pushmeet Kohli said that while scientists have had access to the full human genome since 2003, interpreting it remains a major challenge.
“We have the text,” Kohli said, referring to the three billion nucleotide pairs that form human DNA. “Understanding the grammar of this genome is the next critical frontier.”
Only about two pc of human DNA contains instructions for making proteins. The remaining 98 pc, once dismissed as “junk DNA,” is now known to regulate how genes operate in different cells.
AlphaGenome focuses on this non-coding DNA, which contains many genetic variants linked to disease.
The model was trained using public datasets covering hundreds of human and mouse cell and tissue types. It can analyse long DNA sequences and predict:
- When genes start and stop functioning
- How much RNA is produced
- How genetic changes affect cellular activity
Lead study author Ziga Avsec said the model can process sequences of up to one million DNA letters, allowing scientists to study the full regulatory environment of individual genes.
Potential Impact on Medical Research
Researchers say AlphaGenome could help speed up the identification of genetic factors linked to complex diseases.
Study co-author Natasha Latysheva said the tool can help map functional elements in DNA and explain their molecular roles.
“AlphaGenome can accelerate our understanding of the genome,” she said.
Google said more than 3,000 scientists in 160 countries have already tested the model, which is available for non-commercial research use.
The project is part of Google’s broader AI-driven scientific efforts, which include AlphaFold, the protein-structure tool that contributed to the 2024 Nobel Prize in Chemistry.
Experts believe AlphaGenome could support the development of more targeted therapies by clarifying how specific genetic differences affect disease risk.
Experts Welcome Innovation, Urge Caution
Independent researchers have welcomed the tool while stressing its limitations.
Ben Lehner of the University of Cambridge, who tested the model, said it performed well in identifying genetic differences linked to disease.
“Understanding these differences is key to developing better treatments,” he said.
However, he warned that AI systems depend heavily on training data, which remains incomplete in some areas.
Robert Goldstone, head of genomics at the Francis Crick Institute in the UK, also cautioned against overreliance on the technology.
“Gene expression is influenced by complex environmental factors that the model cannot see,” he said.
Despite these limitations, Goldstone described AlphaGenome as a major breakthrough in studying the genetic roots of complex diseases.
