Researchers at Google DeepMind unveil AlphaGenome, an AI-based software that can usher in a new era in identifying genetic drivers of diseases and eventually pave the way for new treatments. AlphaGenome came into public light in a blog post and a preprint last June. Since then, more than 3,000 scientists from 160 countries have used it, even though access was limited to noncommercial researchers.
Of the 3 billion base pairs— the Gs, Ts, Cs, and As that comprise the DNA code, only 2% tells cells how to make proteins— the coding DNA— while the rest 98%, the non-coding DNA, orchestrates gene activity, and dictates where, when, and how much of individual genes are triggered. We have strong expertise in the 2%, but the remaining 98% is poorly understood and has even been labeled ‘junk DNA’ for being useless. Human Genome Project (HGP) voiced the same. It was not until 2012, when the ENCODE (Encyclopedia of DNA Elements) project showed that ‘junk DNA’ acts as molecular switches and regulates genetic activity.
Various cancers, mental health conditions like schizophrenia, autism, and even Alzheimer’s. Common diseases that run in the family, such as heart disease and autoimmune disorders, are linked to mutations that affect gene regulation. Identifying the genetic glitches responsible is a complex process.
Since the function of non-coding DNA is a relatively recent development in genetics, scientists would like a launchpad to propel research into it. This is where AlphaGenome comes in. It can predict how mutations interfere with gene regulation, in which cells of the body they occur, and whether their biological volume controls are set optimally. The AI can analyze up to 1 million letters of genetic code simultaneously and predict the effects of mutations on various biological processes.
“We see AlphaGenome as a tool for understanding what the functional elements in the genome do, which we hope will accelerate our fundamental understanding of the code of life,” Natasha Latysheva, a DeepMind researcher, said in a press briefing on the work.
The researchers trained AlphaGenome on publicly available human and mouse genome databases to learn associations between mutations in specific tissues and their effects on gene regulation. The DeepMind team believes that the tool will help scientists map which parts of the genome regulate the development of particular tissues, such as nerve and liver tissues, and pinpoint the most important mutations that drive diseases such as cancer. It can also underpin new gene therapies by allowing researchers to design entirely new DNA sequences that can ‘selectively’ switch a certain gene on in one type of cell, but not in another.
Now, scientists aim to perfect the model so that they don’t need to perform an experiment to confirm their predictions. Achieving this goal will require rigorous work from the scientists. AlphaGenome can accelerate CRISPR-based therapies and reduce the risk of “off-target” effects by refining guide RNAs (gRNAs) that ensure the Cas enzyme cuts at the intended DNA sequence.
AlphaGenome joins the field, comprising AlphaFold, which handles protein structure and design, and AlphaMissense, which assesses whether a missense mutation is pathogenic. The triad of AI models can shift the paradigms of genetics and gene therapy.
Copyright @smorescience. All rights reserved. Do not copy, cite, publish, or distribute this content without permission.
