Jeonnam National University Develops AI Technology to Predict Drug Responses in Cancer Patients

Identifying the Most Effective Treatment for Each Patient
Accelerating the Era of Personalized Cancer Therapy

Jeonnam National University’s faculty-founded startup and research team have developed an artificial intelligence (AI) technology that can predict differing drug responses for individual cancer patients. As this opens the possibility of identifying the most effective treatment for each patient in advance, it is expected to accelerate the era of personalized cancer therapy.

Jeonnam National University Campus.

According to Jeonnam National University on April 20, Professor Yoo Seonyong’s team from the College of Engineering and the faculty-founded startup MatiloAI Co., Ltd. have developed an ‘AI technology for precisely predicting transcriptional responses induced by drugs.’

This research was conducted by Chaewon Kim, a researcher in the Department of Intelligent Electronics and Computer Engineering at the College of Engineering, Jeonnam National University. The results have been accepted for publication in the international SCI journal Bioinformatics (Impact Factor 5.4, JCR top 8.7%).

The study addresses technology for AI-based prediction of changes in gene expression when specific drugs are applied to cells. The research team proposed a framework based on a ‘Latent Diffusion Model,’ which combines a Variational Autoencoder (VAE) and a Diffusion Model.

Through this approach, the system simultaneously incorporates various conditions such as cell line, drug structure, treatment concentration, and treatment duration. It has achieved stable predictive performance even for previously unobserved drug-cell combinations. Compared to the previous state-of-the-art (SOTA) model, the new model showed approximately a 7% improvement in Pearson correlation coefficient and reduced computational costs by more than 300 times. Furthermore, experimental validation confirmed that the generated gene expression data faithfully reflect real biological properties.

Drug-induced transcriptional response prediction technology is considered a key enabling technology for virtual screening and drug repurposing in new drug development, where it is difficult to experimentally verify every possible drug-cell combination. In particular, because it can predict individual differences in drug responses by cell type in the field of cancer drug development, it can be directly applied to realizing precision medicine.

Professor Yoo Seonyong stated, “This research is internationally recognized evidence that combining genomic data with generative AI can precisely predict drug responses. In connection with the ‘K-HOPE: Digital Smart Clinical Trial Platform for Korean Cancer Patients’ project, in which Jeonnam National University and MatiloAI Co., Ltd. are participating, we will further develop an AI-driven new drug development platform based on data from Korean patients.”

Meanwhile, Professor Yoo Seonyong founded MatiloAI Co., Ltd., an integrated bio-intelligent platform company that combines bioinformatics data and AI technology, and is working on commercializing the research outcomes.

This research was conducted as part of the Health and Medical Technology R&D Project, funded by the Ministry of Health and Welfare and supported by the Korea Health Industry Development Institute.

This content was produced with the assistance of AI translation services.