Title: Beyond the Blueprint: A Novel Database for Innovative Drug Design and Discovery
Abstract:
Developing a new drug, from identifying a drug target to final marketing, takes 15 years, costs ~$2-5 billion, and has a success rate as low as 2%. To address these challenges, Computer-aided drug design (CADD) has emerged as a promising approach integrating computational methods (e.g., machine learning (ML)) with experimental data to accelerate the drug development process. However, the accuracy of these methods depends on the availability of sufficient training/input data. Many ML drug-discovery models, especially structure-based drug design models, are trained on experimentally determined protein-ligand crystal structures. There is a limited number of databases cataloging these experimental structures (e.g., the Protein Data Bank (PDB) and the Binding MOAD), but there is a vast amount of small-molecule/ligand data without known receptor-bound poses available in other databases (e.g., PubChem). Thus, there is a need for a more comprehensive database of protein-ligand complexes.
I aim to create a database of modeled protein-ligand structures by leveraging the PDB experimental protein structures, the PubChem ligands, and a custom molecular docking framework. I will add PubChem ligands if they share target and structure similarity with PDB structures. Preliminary results have shown that my docking framework can accurately model protein-ligand interactions; however, to further enhance the reliability of the database, I will exclude modeled complexes whose predicted ligand poses differ substantially from related PDB protein-ligand complexes. This database will provide researchers with critical insights into the structural and chemical features that contribute to binding affinity and selectivity, furthering the development of safe and potent drugs.
Durrant Lab
Friday, January 26th, 2023
12:00PM
Langley A219B