Speaker: 申聪，中国科学院数学院应用数学所

Inviter: 苏阳

Title: Higher-Order Topological Learning for Molecular Representation and Drug Response Prediction

Language: Chinese

Time & Venue: 2026.09.14 14:30-15:00 南楼N820

Abstract: Complex molecular systems and biomedical data often contain rich higher-order relational structures. Traditional graph representation learning methods, however, primarily focus on relationships between nodes and edges and therefore struggle to fully characterize the many-body interactions inherent in complex systems. In recent years, higher-order topological structure learning has offered a promising avenue for overcoming the limitations of conventional graph learning by employing richer structural representations to uncover latent organizational patterns in data. This talk focuses on methods for higher-order topological structure learning and their applications to complex networks, molecular representation, and drug response prediction. First, I will introduce deep learning approaches that incorporate topological information and discuss the importance of higher-order structures in enhancing model expressiveness. I will then present applications of higher-order topological learning to molecular systems, including molecular property prediction and structural representation. Finally, I will discuss its use in drug response prediction, highlighting the potential of higher-order structural modeling for biomedical data analysis. Overall, this talk aims to demonstrate how higher-order topological structure learning can provide new intelligent approaches for analyzing complex molecular and biological systems, while exploring its broader prospects in AI-assisted scientific research.