Bidirectional path-based nonequilibrium simulations for binding free energy estimation

Drug discovery is an inherently complex, time-consuming, and costly process, often requiring over a decade of research and investments amounting to billions of dollars. A key quantity that underpins the success of early-stage drug design is the binding free energy, which quantifies the strength of interactions between a protein and a potential ligand. Accurate estimation of binding free energy is essential for identifying promising drug candidates and prioritizing them for experimental validation. This thesis first gives a comprehensive examination of the theoretical principles that govern simulation for binding free energy estimation, and then introduces novel protocols. We give a detailed review of the statistical mechanics framework for binding free energy calculations and critically evaluate key free energy estimators by analyzing their underlying assumptions, practical advantages, and limitations. In addition to consolidating theoretical knowledge, this work contributes practical insights by assessing the applicability and the performance of some of these estimators, particularly nonequilibrium estimators, across representative molecular systems. The findings support and clarify the use of nonequilibrium approaches under diverse conditions and in increasingly complex systems, while also providing guidance for best practices in real-world drug discovery pipelines. Taken together, these contributions strengthen the theoretical foundations of protein-ligand binding free energy estimation and give valuable tools and guidelines to improve the efficiency of computational drug discovery efforts.