UNMC researchers build AI tool that predicts breast cancer subtype from DNA mutation data
The research tool, called ViSTA, is not making clinical claims yet, the team says.
OMAHA - Researchers at the University of Nebraska Medical Center have built an artificial intelligence tool that predicts a breast cancer's molecular subtype using only DNA mutation data, according to a UNMC news release issued Tuesday. The release says the study was published in September in the journal Genome Biology.
Breast cancer has several subtypes that behave and respond to treatment differently. Subtyping has usually relied on gene expression data, and combining several kinds of patient data can be costly, the release says.
The tool, ViSTA, adapts a Google language-model method called BERT, treating DNA sequences as a language and mutation patterns as words. To make the work manageable, it looks only at the DNA around each mutation. Training took about three weeks per round, the release says.
We use only mutation data, which is very cheap to obtain, relatively.
Guda led the study, and Sushil Shakyawar, PhD, an instructor in UNMC's Department of Genetics, Cell Biology and Anatomy, is first author. The team tested ViSTA on two independent breast cancer cohorts and made the model public for other researchers. It remains a research tool, and the team is not making clinical claims yet, according to the release.
Source: UNMC Newsroom, "Researchers develop tool to predict breast cancer subtypes," Nov. 3, 2026.