DeepMind’s AlphaGenome AI Decodes ‘Dark Genome’ & Accelerates Drug Discovery

A new artificial intelligence model developed by Google’s DeepMind is poised to significantly advance our understanding of DNA, potentially unlocking solutions to long-standing medical mysteries. This breakthrough follows DeepMind’s success with AlphaFold, which earned the team the 2024 Nobel Prize in Chemistry for its ability to predict protein structures.

Unraveling the “Dark Genome”

What is AlphaGenome?

The new AI, named AlphaGenome, is being hailed as a “remarkable achievement” and a “major milestone” by experts. Its primary function is to help scientists understand why even small differences in DNA can increase the risk of diseases like high blood pressure, dementia, and obesity. DeepMind research engineer Natasha Latysheva describes AlphaGenome as a tool to understand the function of elements within the genome, with the goal of accelerating our fundamental understanding of life’s code.

Did You Know? The human genome is comprised of 3 billion DNA “letters” (A, C, G, T), with only about 2% coding for proteins.

Approximately 98% of the human genome has historically been poorly understood, referred to as the “dark genome.” This region plays a crucial role in organizing how genes function, and many disease-related mutations are located within it.

How Does AlphaGenome Work?

AlphaGenome demonstrates powerful computational capabilities, analysing up to one million DNA letters at a time. It can predict gene locations, analyze the influence of the dark genome – including gene expression levels and the process of gene splicing – and even predict the consequences of a single letter change in the genetic code. Natasha Latysheva notes the model’s potential to illuminate the mechanisms behind disease and identify causes of rare genetic conditions, ultimately aiding in drug discovery and the development of new therapies.

The findings related to AlphaGenome have been published in the journal Nature and the model was made available for non-commercial use last year, and has since been utilized by over 3000 scientists. Gareth Hawkes of the University of Exeter is using the tool to investigate how mutations impact the risk of obesity and diabetes. He explains that AlphaGenome can quickly predict the biological effects of variations often found within the dark genome, guiding further laboratory testing and drug development.

Potential Applications in Cancer Research

AlphaGenome is also showing promise in cancer research, helping to distinguish between mutations that drive cancer growth and those that are merely incidental. Robert Goldstone, head of genomics at the Francis Crick Institute, describes the model as a significant advancement in genomic AI, capable of predicting gene expression based solely on DNA sequence.

Expert Insight: The ability to rapidly analyze and interpret the vast, previously uncharacterized “dark genome” represents a paradigm shift in genetic research, potentially accelerating the pace of discovery for a wide range of diseases.

Researchers at the Wellcome Sanger Institute have conducted over 500,000 experimental tests, demonstrating AlphaGenome’s strong performance. While the developers acknowledge the model isn’t perfect – with ongoing improvements needed for predicting long-distance gene regulation and differences between tissues like brain cells and heart cells – it signifies a new era of scientific progress.

A Different Approach to AI

AlphaGenome differs from large language models like ChatGPT, which predict the next word in a sequence. Instead, it employs a “sequence-to-function” approach, focusing on how changes in the DNA sequence impact biological outcomes. The model was trained using publicly available data from human and mouse cell experiments.

Pushmeet Kohli, VP of Science and Strategic programmes at Google DeepMind, emphasizes that humanity is at the beginning of a new era of scientific advancement, with AI playing a crucial role in driving breakthroughs. He suggests that the convergence of genomics, biomedical research, and AI in the UK could lead to transformative changes in biology and medicine.

Frequently Asked Questions

What is the “dark genome”?

The “dark genome” refers to the approximately 98% of the human genome that does not code for proteins. Despite not directly coding for proteins, it plays a critical role in regulating gene function and is often the location of disease-related mutations.

What is AlphaFold’s connection to AlphaGenome?

AlphaGenome builds upon the success of DeepMind’s AlphaFold, which predicted protein structures and earned the 2024 Nobel Prize in Chemistry. Both models demonstrate DeepMind’s commitment to using AI to solve complex biological problems.

How is AlphaGenome being used currently?

AlphaGenome is currently being used by over 3000 scientists for non-commercial research, including exploring the genetic basis of obesity, diabetes, and cancer. Researchers are using it to predict the impact of genetic variations and identify potential drug targets.

As AlphaGenome continues to develop, what further insights into the complexities of the human genome might it reveal?