As AI continues to advance across various fields, its application in identifying disease-related genes is expanding. While traditional AI systems only predicted gene-disease associations, a research team at Pusan National University has developed an AI system that also determines whether a gene can serve as a therapeutic target or biomarker. This breakthrough is expected to accelerate precision medicine by enabling gene-based treatments tailored to individual patients.

Led by Professor Giltae Song in collaboration with Professor Hyewon Lee’s team at Pusan National University Hospital, the system uses hypergraph structures and attention mechanisms to analyze complex biological interactions and provide interpretable results. Validated through curated open-source biological databases like DisGeNET, the model enhances the accuracy of identifying therapeutic gene candidates.

Professor Song emphasized that this research surpasses conventional methods by refining gene predictions, ultimately facilitating faster drug development and targeted treatments. The first author of the study, Kibeom Kim, a Ph.D candidate researcher in AI and a student of Professor Song, played a key role in its development. The team has filed a domestic patent and is pursuing a U.S. patent with university support. The study, backed by major Korean research institutions, was published in Briefings in Bioinformatics on January 22, marking a significant step in AI-driven medical innovation.

- Authors (Pusan National University)

 · First author: Kibeom Kim (Division of Artificial Intelligence)

 · Corresponding author: Giltae Song (Department of Electrical and Computer Engineering)

- Title of original paper: Therapeutic gene target prediction using novel deep hypergraph representation learning

- Journal: Briefings in Bioinformatics

- DOI: https://doi.org/10.1093/bib/bbaf019