学术报告会：Deep Learning Approaches for DiseasePrediction, Toxicity Assessment, and DrugRepurposing

报告题目：Deep Learning Approaches for DiseasePrediction, Toxicity Assessment, and DrugRepurposing

报告人：何岱海（香港理工大学）

报告时间：2025年5月18日周日 9:00

报告地点：S3-313

摘要：This study introduces an AI-driven framework that combines deep learning with topological data analysis (TDA) to address key challenges in thebiomedical field, including prediction of chronic kidney disease (CKD)progression, repurposing of Alzheimer’s disease (AD) medications, andassessing the toxicity of PM2.5 components. By leveraging PersistentLaplacians and deep learning algorithms, these models improve predictiveaccuracy and interpretability for a variety of biomedical applications. Deeplearning models of CKD progression provide valuable insights into earlyintervention strategies and provide diabetic patients with a platform that canbe relied upon for predicting the risk of progression over different durationsof chronic kidney disease. For AD drug repurposing, the fusion of TDAand deep learning successfully identified promising drug candidates, butfurther validation is needed. PM2.5 toxicity prediction models demonstratedthe potential of topological descriptors to improve environmental healthassessment, although inconsistencies in dataset performance suggest that themodels and data require additional optimization. The findings emphasize thefeasibility of combining TDA and AI for biomedical applications and outlinefuture directions for improving model robustness and generalizability.

个人简介：

何岱海，香港理工大学应用数学系教授，博士生导师。分别于1999年获西安交通大学工学博士和2006年加拿大麦克马斯大学数学博士，并且曾在北京师范大学物理系、美国密西根大学生态学系、以色列特拉维夫大学动物学系做博士后研究。主要研究兴趣是传染病建模和数据统计分析，在PNAS, Sci Adv, Ann Intern Med, Eur Respir J, J R Soc Interface等权威期刊发表论文140余篇，研究成果受到国内外媒体的广泛报道。关于非洲安哥拉黄热病的建模获2018年国际疾病监测学会的科学贡献最佳论文第二名；先后获得香港研究资助局项目、香港食品与卫生环境署健康与医疗项目、阿里巴巴合作研究基金等多项基金资助。Google H-index 47.连续三年入选斯坦福大学发布的全球2%顶尖科学家榜单(2022-2024)，以及ScholarGPS2024全球前0.05%学者。