AI maps nine billion possible DNA changes

Google DeepMind has created an artificial intelligence based atlas that predicts how single letter changes in human DNA may affect genes and other cellular processes, ERR reports.

Called AlphaGenome, the genome atlas contains predictions for around nine billion possible changes. The model analyses DNA sequences to estimate how a change could affect gene activity and other processes inside cells.

The tool is particularly important for studying the noncoding parts of the genome, which do not directly provide instructions for making proteins. These regions account for about 98 percent of human DNA and include sequences that influence when and how actively genes work. Scientists still do not fully understand the function of many of these regions.

According to the researchers, AlphaGenome can analyse sequences containing up to one million DNA letters at a time. Because the effects of possible changes have been calculated in advance, scientists can search and compare the results quickly.

The tool could help researchers identify genetic changes that warrant closer examination when investigating the causes of diseases. However, its predictions do not prove that a specific change causes disease and cannot replace laboratory tests or medical assessments.

The study introducing the new genome atlas has not yet undergone peer review. Further information about the AlphaGenome model is available in Nature.