New AI tool predicts which diseases human DNA can cause

Researchers at the Icahn School of Medicine at Mount Sinai Hospital have developed a new AI tool that can not only detect genetic changes associated with disease, but also predict which pathologies these mutations might cause. The technology, called V2P (Variant to Phenotype), is designed to speed up genetic diagnostic processes and facilitate the development of new approaches to the treatment of complex and rare diseases. The results of the study were published in the online version of the journal Nature Communications.

Most existing genetic analysis tools are able to determine whether a particular mutation is dangerous, but, as a rule, they are not able to pinpoint exactly what kind of disease it can cause. V2P addresses this limitation by using advanced machine learning algorithms to associate genetic variants with their likely phenotypic manifestations—that is, diseases or traits that may result from a mutation. Thus, the system actually predicts the influence of a person’s DNA characteristics on his health. To train V2P, the researchers used a rich data set that included both pathogenic and neutral genetic changes, as well as detailed information about the diseases with which they are associated. When tested on real anonymized patient data, the algorithm in many cases placed the true causative mutation in the top 10 most likely variants, indicating the technology’s potential to significantly speed up and simplify the genetic diagnosis process.

Currently, V2P classifies mutations into large classes of diseases, for example, relating them to pathologies of the nervous system or cancer. Going forward, the researchers aim to improve the tool’s accuracy so it can predict more specific clinical outcomes, and integrate it with additional data sets to enhance research and development of new drugs.

This breakthrough takes another step toward the development of personalized medicine, in which therapy is tailored to a patient’s individual genetic makeup. By matching specific genetic changes with their potential impact on health, V2P can speed up the diagnostic process for doctors and help researchers find promising new avenues for developing treatments, the scientists note.