UVA Researchers Unveil AI Tool for Mutation Analysis

The deep-learning tool, called ERIS2, predicts how genetic mutations affect protein stability and is free for scientists to use.

Updated on Nov. 9, 2026 in Biotech

Translucent folded protein-like sculpture on a white laboratory bench beside a window in soft daylight, with a blurred lab interior behind it. — University of Virginia researchers led by Nikolay Dokholyan released ERIS2, a free deep-learning tool that predicts how genetic mutations affect protein stability. AI Illustration. Upload story photo >

Live Poll

Do you think AI tools speeding up medical research will benefit you more than they risk harm?

University of Virginia researchers led by Nikolay Dokholyan, PhD, of the department of neurology have developed ERIS2, a deep-learning AI tool that predicts how genetic mutations affect protein stability. The findings were published in the journal Structure, and the tool is available free to the scientific community without registration.

Why it matters

Understanding how mutations alter protein stability is central to studying disease and drug development, and a freely available predictive tool can speed that work for researchers. Because ERIS2 was built on a large body of known mutations, it offers a data-driven way to screen variants before costly laboratory testing.

ERIS2 was trained on a database of more than 100,000 known mutations to predict changes in protein stability. The source does not report the model's accuracy or how it compares with other predictive tools.

The players

Nikolay Dokholyan, PhD

Dokholyan is a researcher in the University of Virginia's department of neurology who led the development of the ERIS2 protein-stability tool.

University of Virginia

The University of Virginia is a public research university whose School of Medicine in Charlottesville, Virginia, produced the ERIS2 research.

Structure

Structure is a peer-reviewed structural biology journal in which the ERIS2 findings were published.

The details

ERIS2 was developed by researchers led by Nikolay Dokholyan, PhD, of UVA's department of neurology, and is a deep-learning tool that predicts how genetic mutations affect protein stability. The team published its findings in the journal Structure and is sharing the tool free with the scientific community without requiring registration.

Nov. 9, 2026: UVA announced ERIS2 in a news release, which was shared with Becker's.

The Tech Race

ERIS2 extends the deep-learning approach to protein biology popularized by AlphaFold, shifting focus from predicting static protein structures to forecasting how individual mutations change protein stability. The tool joins a growing set of AI systems aimed at turning large biological datasets into predictive models for researchers.

Researchers studying genetic variants can use ERIS2 at no cost to screen how mutations may affect protein stability before running lab experiments. The tool's free, no-registration access lowers the barrier for smaller labs that lack commercial software budgets.

The takeaway

Free, openly available AI tools for predicting mutation effects can let more researchers test ideas quickly, so it is worth watching whether ERIS2 is adopted and validated by labs beyond its developers.