Novel Tool Finds Genetic Variations That Predict Disease
REVEL developers Weiva Sieh, MD, PhD, left, and Joseph Rothstein, MS
Sequencing-based genetic tests are making personalized medicine a reality for patients who carry certain mutations for serious diseases, including breast and prostate cancer. Armed with information about the pathogenicity of such variants in their genome, these patients can now make proactive decisions about their health care early enough to increase the odds of success in preventing or treating disease.
But for each step forward in this new frontier of personalized medicine, physicians and scientists are stymied by their lack of knowledge about thousands of other rare variants in a patient’s genome that could also portend disease. For this reason, two researchers at the Icahn School of Medicine at Mount Sinai developed REVEL, a novel tool designed to make sense of these variants of unknown importance in order to help guide patient care and risk management and to facilitate research.
Developed by Weiva Sieh, MD, PhD, Associate Professor of Population Health Science and Policy, and Genetics and Genomic Sciences; and Joseph Rothstein, MS, an Instructor of Genetics and Genomic Sciences, and Population Health Science and Policy, REVEL is an acronym for Rare Exome Variant Ensemble Learner. The tool predicts the likelihood of whether a particular coding variant in a person’s genome is disease-causing or benign.
“REVEL is timely and significant because the number of rare variants discovered by sequencing studies is vast and growing and little is known regarding their function,” says Dr. Sieh. “Yet few pathogenicity prediction tools have targeted rare missense variants, leaving researchers and clinicians to struggle with their interpretation.” The ability to distinguish among the approximately 10,000 missense variants in each person’s genome that result in protein changes that could potentially be harmful also helps researchers set priorities for the variants they may wish to study further as new disease-causing genes. Identifying more of these variants would enable personalized medicine to realize its potential in helping more patients benefit from prevention and treatment regimens tailored to their individual disease risks.
Other tools that use different predictive features do exist, but they often do not agree with each other on the likelihood of pathogenicity. Dr. Sieh and Mr. Rothstein created REVEL as an ensemble method that uses machine learning to combine information from many of these other tools in order to generate a consensus score. REVEL incorporates 18 pathogenicity prediction scores from 13 tools, and it specializes in prioritizing the most clinically or functionally relevant variants among the sea of rare variants that are being discovered through next-generation sequencing.