Dr. Alexander Tropsha
Dr. Alexander Tropsha
The University of North Carolina, Chapel Hill
Associate Dean for Pharmacoinformatics and Data Science, UNC Eshelman School of Pharmacy
Dr. Alexander Tropsha is K.H. Lee Distinguished Professor and Associate Dean for Pharmacoinformatics and Data Science at the UNC Eshelman School of Pharmacy.

Dr. Tropsha's research interests are in the areas of Computational drug discovery, Cheminformatics, and Structural Bioinformatics. He has authored more than 230 peer-reviewed research papers, reviews and book chapters.

Topic & Abstract

Curated Data + Rigorous Methods = Predictive Science

Dr. Tropsha will discuss several topics of relevance to enabling accurate predictions of chemical bioactivity: (i) data curation; (ii) methods, models, and approaches to bioactivity prediction and elucidation of the mechanisms of action (MOA) of bioactive compounds; (iii) approaches to de novo design of new chemicals with the desired properties. Dr. Tropsha will briefly discuss the dependency of the model accuracy on the input data quality using case studies of chemical toxicity prediction. To illustrate best practices for data curation and model development, Dr. Tropsha will introduce the recently developed STopTox (Systemic and Topical Toxicity) software package for evaluating putative toxicity of chemicals tested in the well-known regulatory “6-pack” assays. He will then describe the Biomedical Data Translator and Reasoning project funded by the NIH. In this project, they integrate data in multiple biomedical databases, or knowledge sources, into a comprehensive Knowledge Graph where individual biomedical entities such as drugs, biological targets of drug action, and diseases form nodes, and functional relationships between these entities are encoded as edges. Dr. Tropsha will discuss reasoning over the Knowledge Graph for uncovering possible mechanistic relationships between distinct entities such as chemicals and their effects in vivo. Finally, Dr. Tropsha will present a novel Artificial Intelligence and deep learning based computational strategy for de-novo design of molecules with the desired properties termed ReLeaSE (Reinforcement Learning for Structure Evolution). He will describe the methodology and present the initial results of using ReLeaSE for designing novel compounds that showed the desired experimental activity. In summary, Dr. Tropsha will describe computational tools and methods that help design experimental drug discovery studies and support regulatory decisions concerning chemical safety.