Combining deep learning and molecular dynamics in computer-aided drug discovery

Computer-Aided Drug Discovery (CADD) is a fast-evolving, interdisciplinary field that incorporates diverse aspects of computational chemistry, medicinal chemistry, physics, and computer science to develop new drugs or optimize existing ones. CADD has become indispensable to the process of drug discovery as a tool that can significantly reduce the time and cost involved, and many efforts are being made to constantly improve these in silico methods and techniques. A large part of CADD is based on modeling and molecular dynamics simulations, calculating the movement of molecules over time with algorithms. This dissertation is a collection of studies that combine deep learning with molecular dynamics in various approaches. The first chapter provides a general introduction to the various topics covered throughout the thesis, including an overview of developments in deep learning-based drug discovery techniques, descriptions of deep learning neural networks, and theoretical background for free energy calculations. The second chapter addresses how deep learning can be used to improve the targeted free energy perturbation method for free energy difference estimations of biomolecules. The third chapter investigates deep learning methods for conformation sampling of biomolecules of large sizes to address the sampling problem in computational chemistry. The fourth chapter discusses the use of deep learning in sampling when combined with coarse-grained simulation data, and presents a hierarchical approach that is better equipped to meet challenges of the sampling problem. The final chapter gives concluding remarks regarding the collective works in the dissertation and possible future directions for further research. The aim of this dissertation is to present findings from studies that combine artificial intelligence with conventionally used, molecular dynamics-based methods in CADD and contribute new methods and approaches for in silico drug discovery.