UVA scientists unveil free AI tool to predict how gene mutations affect proteins

CHARLOTTESVILLE, Va. — Researchers at the University of Virginia School of Medicine have built an artificial intelligence tool designed to predict how genetic mutations change the stability of proteins, work the university says could speed the search for treatments for inherited disorders and cancers.

The deep-learning tool, called ERIS2, was developed by a team led by Nikolay V. Dokholyan of UVA’s Department of Neurology, UVA Health announced Monday. The researchers described the tool in a new article in the scientific journal Structure.

A protein’s stability matters because its shape governs how well it performs its jobs inside cells, such as binding to enzymes. Mutations that destabilize a protein can lead to disease.

“Understanding how mutations affect protein stability is key to understanding protein function and often disease,” Dokholyan said. “Predicting how mutations change protein stability helps us understand protein function and its relationship to disease, while also enabling us to design new proteins that can serve as biotechnological, diagnostic or therapeutic tools.”

According to UVA Health, current laboratory methods for measuring those effects can be costly and labor-intensive, and are hard to apply to long chains of amino acids, the building blocks of proteins. The team trained ERIS2 on a database of more than 100,000 known mutations, combined with detailed protein measurements and structural information.

In testing, the researchers found the tool offered significantly improved accuracy in predicting the effects of mutations, the university said. Dokholyan has made ERIS2 available to scientists for free online, with no login or registration required.

The work ties into the mission of UVA’s new Paul and Diane Manning Institute of Biotechnology, which aims to move scientific discoveries into medicines more quickly.

The research team included Srinivasan Ekambaram, Debarshi Kundu, Swaroop Ghosh and Dokholyan, who have no financial interest in the work, according to the release. Development of ERIS2 was supported by National Science Foundation grants and gifts from Intel. Dokholyan has also received support from the National Institutes of Health.

This report is based on a press release from UVA Health.