Receive free artificial intelligence updates

Researchers at Google DeepMind used artificial intelligence to predict whether human genetic mutations might be harmful, one of the first examples of the technology helping to speed up diagnosis of diseases caused by genetic variants.

An artificial intelligence tool called AlphaMissense evaluated all 71 million “missense” mutations, in which a single letter of the human genetic code is changed. Of these, 32% were classified as probably pathogenic, 57% as benign, and the remainder indeterminate.

The findings were published Tuesday in the journal science.

So far, human experts have discovered only 0.1% of these mutations, which alter the structure of proteins, the body’s main working molecules, have clinical effects. “Experiments to discover disease-causing mutations are expensive and laborious,” says Žiga Avsec, a researcher at DeepMind’s London-based program.

“Each protein is unique and each experiment must be designed individually, which can take months,” Avsec said. “By using AI predictions, researchers can preview results for thousands of proteins at once, which helps prioritize resources and accelerate more complex studies.”

“We should emphasize that these predictions were never really intended to be used solely for clinical diagnosis,” said Jun Cheng, a researcher on the project. “They should always be used in conjunction with other evidence. However, we do think our predictions will help improve the diagnosis of rare diseases and potentially help us find new disease-causing genes.”

AlphaMissense predictions showing mutations in two protein structures (see additional images).Red is harmful, blue is benign, and gray is uncertain

Ellen Thomas, the UK government’s deputy chief medical officer for Genomics England, said the tool’s predictions were compared with the tool’s extensive record of genetic variants that cause rare diseases, and she was impressed by the results. profound.

“We were not involved in generating the tool or providing the data to train it, so we can conduct an independent evaluation,” Thomas said. “It’s completely different than the tools we’re already using. I think it’s a huge improvement and we’re excited to be involved in the final stages of considering the tool.”

Thomas said she expects AlphaMissense to be used in health care as “a co-pilot for clinical scientists, flagging which variants they should focus on so they can do their jobs more effectively.”

DeepMind developed AlphaMissense based on its AlphaFold tool for predicting protein structures. The AI ​​tool also draws from a wealth of biological evidence to understand the signatures of mutations that make genetic variants pathogenic or benign in humans and other primates.

The company, founded in 2010 as a professional artificial intelligence developer and acquired by Google in 2014, has made the tool “free to the scientific community.” Its predictions will be incorporated into the widely used Ensembl Variant Effect Predictor run by the European Bioinformatics Institute in Cambridge.

AlphaMissense has its limitations, Avsec said. Most importantly, its predictions of pathogenicity “are made in a general sense and do not tell us about the biophysical properties of the variants”. He added that these insights may emerge more clearly as the tool is further developed.

Sarah Teichmann, head of cytogenetics at the Wellcome Sanger Institute in Cambridge, who was not involved in the study, said that while individual missense mutations are important causes of disease, other clinically significant changes in DNA are beyond the scope of the tool.

“We shouldn’t exaggerate and say this is going to solve everything,” she said. “But having such powerful interpretive artificial intelligence integrate so much genomic data is a real advance.”


Leave a Reply

Your email address will not be published. Required fields are marked *