Busan National University Finds Disease Genes and Analyzes Biomarker Potential Using AI

Joint Research Achievement by Professor Song Giltae’s Team at Pusan National University and Professor Lee Hyewon’s Team at Pusan National University Hospital, Patent Application Filed
Development of Technology for Predicting Disease Therapeutic Targets and Biomarkers Based on Open-Source Ontology Data

As AI is applied across various fields, it is also being utilized in the discovery of disease-associated genes.

While existing AI systems merely predicted the association of specific genes with diseases, a research team at Pusan National University has succeeded in developing an AI system that simultaneously predicts the disease association of genes and whether those genes can function as therapeutic targets or biomarker genes for diseases. There is growing anticipation that this will accelerate the era of precision medicine, providing patient-specific gene-tailored treatments through AI.

The research team led by Professor Song Gil-tae from the Department of Information and Computer Engineering at Pusan National University (President Choi Jae-won), in collaboration with Professor Lee Hye-won from the Department of Cardiology at Pusan National University Hospital, announced on the 12th that they have developed an AI system that predicts whether genes are therapeutic gene targets or biomarker genes for diseases and provides sufficient explanations for the results.

From the left, Professor Song Gil-tae, Professor Lee Hye-won, Researcher Kim Ki-beom. Provided by Pusan National University

Diseases arise from the complex interactions of multiple genetic factors within an individual. The research team proposed an AI system that reflects these interactions to predict the potential of certain genes as therapeutic genes or biomarker genes. This system models the complex interactions among various biological factors involved in diseases using hypergraph and attention mechanisms, and provides explanations for the model’s predictions through visualization of the attention operation results.

A hypergraph is a type of network data defined as a set of nodes and hyperedges, where each hyperedge connects two or more nodes simultaneously, capturing complex and shared interactions among them.

Attention is a type of computational algorithm widely used in state-of-the-art artificial intelligence (deep learning) models. It mimics human ‘focus’ by enabling AI models to concentrate more on important information and less on relatively less important information. Visualizing the results of the model’s attention mechanism operation reveals which information the model considered more important in making specific decisions.

The research team validated the developed AI system using relationship data among genes, gene ontology (a network revealing conceptual relationships expressed in language), diseases, disease ontology, and human phenotype ontology obtained from open-source databases curated by biological experts, including DisGeNET.

Professor Song Gil-tae of Pusan National University stated, “This research is significant in that it goes beyond previous studies that simply predicted the association between diseases and genes, by developing a practical AI system that precisely predicts the potential of specific genes as therapeutic genes and biomarker genes.” He added, “By utilizing the proposed AI system, therapeutic gene candidates for specific diseases can be discovered in a short time, bringing us one step closer to realizing precision medicine that directly targets disease-causing genes to eliminate the root causes of diseases.”

The research team has completed domestic patent applications for the AI system and is currently pursuing a U.S. patent application with support from Pusan National University’s Industry-Academic Cooperation Foundation.

This research was supported by the Basic Research Laboratory and Mid-career Research Projects of the Ministry of Science and ICT and the National Research Foundation of Korea, as well as the AI Convergence Innovation Talent Training Project supported by the Institute for Information & Communications Technology Planning & Evaluation. Professor Song Gil-tae of the Department of Information and Computer Engineering at Pusan National University served as the corresponding author, PhD candidate Kim Ki-beom from the AI major at Pusan National University Graduate School was the first author, and Professor Lee Hye-won from Pusan National University Hospital was a co-author.

The research results, expected to overcome the limitations of existing AI systems in predicting associations between diseases and genes, were published on January 22 in the international journal Briefings in Bioinformatics, published by the University of Oxford in the United Kingdom.