Google DeepMind’s AlphaGenome Atlas could help researchers find the genetic causes of diseases - Inquirer Technology

AlphaGenome Atlas contains predictions for the molecular effects of all roughly 9 billion possible single-letter changes in human DNA. Researchers can search the database to see how individual variants are predicted to affect processes such as gene expression, RNA splicing, and chromatin accessibility across different cell types and tissues.

That gives researchers something they have not previously had at this scale: a precomputed view of how individual changes across the genome could affect the machinery that regulates genes.

Testing billions of possible variants experimentally is not realistic, particularly when researchers are trying to identify the cause of a rare disease. AlphaGenome Atlas instead allows researchers to narrow the field before taking the most interesting candidates into laboratory work.

Researchers from the Broad Institute and the GREGoR Consortium used AlphaGenome’s Variant Impact score to prioritize possible disease-causing variants. The system identified a variant affecting DNM1, a gene associated with epileptic encephalopathy, and predicted that the mutation would create an incorrect splice site. Laboratory experiments subsequently supported the prediction and found nearby variants with similar effects.

The value of the system in this case was not that an AI diagnosed the disease. The researchers still needed laboratory experiments to establish what the variant was doing. AlphaGenome helped narrow down where they should look.

The platform is a research tool, not a clinical diagnostic system. Google DeepMind says AlphaGenome has not been validated or approved for clinical use, so its predictions cannot currently be treated as medical diagnoses or used on their own to determine treatment.

AlphaGenome Atlas assigns each variant an AlphaGenome Variant Impact, or AVI, score that combines predictions from AlphaGenome and AlphaMissense. Researchers can use the score to rank variants and then examine which biological processes are predicted to be affected. The Atlas also connects those predictions to specific regulatory features, giving researchers more information about why a particular variant received its score.

Researchers at the University of Exeter used AlphaGenome Atlas with whole-genome data from more than 54,000 UK Biobank participants to investigate rare non-coding variants associated with human traits. By grouping variants according to their predicted molecular effects, the researchers identified 22% more non-coding genetic associations than they otherwise detected.

That kind of analysis could be useful for studying diseases and traits where the genetic picture is more complicated than a single mutation causing a single condition. The same underlying problem remains: researchers have to distinguish meaningful genetic signals from a huge amount of variation that may have little biological effect.

AlphaGenome Atlas also gives researchers a way to look at how regulatory DNA behaves across different cells and tissues. A DNA sequence can be present throughout the body while producing very different effects depending on which genes a particular cell activates. Understanding those differences is central to studying how genetic changes contribute to disease.

A genetic variant that looks insignificant when viewed simply as a change in DNA letters may have a very different interpretation once researchers can see that it is predicted to interfere with gene regulation in a particular type of cell. That information can give scientists a more specific question to take into the laboratory.

AlphaGenome is already being used in research beyond rare disease. Google DeepMind says researchers have used it to examine genetic signals associated with Alzheimer’s disease, including variants whose effects may operate through immune and metabolic tissues rather than only in the brain.

There is still a considerable distance between a genetic prediction and a medical treatment. Laboratory validation remains necessary, and diseases are influenced by biological processes that cannot always be inferred from DNA sequence alone.

For now, AlphaGenome Atlas is best understood as a way of making the genome easier to investigate.

Researchers already have the sequence. What has been harder is determining which changes matter, what those changes do inside cells, and which ones deserve closer attention.

AlphaGenome Atlas does not answer all of those questions. It gives researchers a way to search through them faster, with the aim of finding the variants that warrant the next experiment.