AlphaGenome: Google’s AI Predicts DNA Behavior & Disease Risk

Our DNA, comprised of millions of building blocks, orchestrates the functions of the human body. Even slight alterations in these sequences, or how they operate, can disrupt bodily processes and contribute to diseases like cancer. Now, Google has introduced a new artificial intelligence tool, AlphaGenome, designed to decipher vast stretches of DNA and predict how different segments behave and potentially lead to illness. The tool leverages deep learning, inspired by the way the human brain processes information, to aid scientists in understanding the complexities of the genome.

How AlphaGenome Works

AlphaGenome can accurately read up to 1,000,000 characters of DNA, a feat previously unattainable. DNA is constructed from long chains built upon four core chemical building blocks – nucleotides – represented by the letters “A,” “C,” “G,” and “T.” These act as a blueprint for creating and controlling every cell. While only about two percent of human DNA directly codes for proteins, the remaining 98 percent—often previously dismissed as “junk DNA”—regulates the activity of those protein-coding regions.

This 98 percent adjusts when, where, and how strongly genes are turned on or off, responds to environmental signals, and influences the creation of RNA. Many disease-related variations are hidden within this regulatory region, impacting gene activity without altering the proteins themselves. AlphaGenome is the first deep learning model specifically designed to target and predict the function of this portion of the genome, and to estimate the impact of small genetic changes, known as variants, on gene activity or the disruption of natural processes linked to diseases like cancer.

Practical Applications of AlphaGenome

Researchers focused on a form of acute leukemia, a cancer affecting white blood cells when immature T-cells grow uncontrollably, as a real-world example. Some leukemia cases are linked to small changes in DNA that don’t alter the protein, but modify the timing or intensity of gene expression. AlphaGenome compared a normal DNA sequence to a mutated version, predicting the likelihood that the mutation would increase activity in nearby genes.

Currently, the model is available to scientists free of charge for non-commercial research purposes and is not intended for clinical use.

Potential Benefits and Future Directions

Researchers envision multiple uses for the new model. In molecular biology, it can serve as a virtual laboratory tool, allowing for hypothesis testing through simulation before conducting costly experiments. In biotechnology, it could aid in designing gene therapies or improving molecules that target specific tissues. Robert Goldston, head of the Genome Biology department at the Francis Crick Institute, stated that “AlphaGenome from DeepMind represents a major milestone in the field of AI for genomics,” adding that the model’s accuracy represents a breakthrough moving the technology from theoretical interest to practical application.

Goldston further explained that AlphaGenome “isn’t a magic bullet for all biological questions,” but is a high-quality tool that transforms the static code of the genome into a decipherable language.

However, scientists caution that AlphaGenome, like all AI models, is only as good as the data it’s trained on. Ben Lehner, head of the Generative and Synthetic Genomics department at the Wellcome Sanger Institute, noted that “most of the data available today in biology is not very suitable for AI,” as datasets are often too small and poorly standardized. He believes the primary challenge now is producing the necessary data to train the next generation of AI models.

Frequently Asked Questions

What is AlphaGenome?

AlphaGenome is a new artificial intelligence tool developed by Google capable of reading large sections of DNA and predicting how changes in the genome might lead to disease.

What percentage of human DNA does AlphaGenome focus on?

AlphaGenome focuses on the 98 percent of human DNA that does not directly code for proteins, but instead regulates gene activity.

Is AlphaGenome available for clinical use?

No, AlphaGenome is currently a research tool and is not intended for clinical use.

As AI tools like AlphaGenome continue to evolve, what role do you foresee them playing in personalized medicine and disease prevention?