The early stages of drug design rely on hit discovery programs, where initial possible inhibitors’ binding affinities are assessed when bound to their biological target. It is an expensive and time-consuming process, requiring multiple iterations of trial and error designs. This sets the perfect ground for computer simulations. Structure-based drug design has been in the past decade a widely used computational methodology to speed up the drug discovery process for resolved protein-ligand systems[1]. However, providing a fast and reliable answer to the protein-ligand affinity problem can be an arduous task. In this context, the capacity of the software to score the binding affinity of the inhibitors will be crucial to determine possible drug leads that will be later on optimized. Hence, the main goal of this research is to add physically justified corrections as well as Machine Learning models to the energetic predictions to obtain absolute binding free energies that match the experimental results. To do it we will need to review the physics involved in the forcefields used in the simulations done with the software used in the group: PELE[2]. PELE stands for Protein Energy Landscape Exploration and it is a self-contained Monte Carlo software to model protein-ligand interactions. The reachable conformations by the protein and ligand are explored and energetically assessed with the forcefield. The forcefield is the parameterized functional (eq. 1) that enables a Monte Carlo or a Molecular dynamics simulation to calculate the potential energies involved[3]. Etotal = Ebonded + Enonbonded Ebonded = Ebond + Eangle + Edihedral Enonbonded = Eelectrostatic + Evan der Waals. (1) This functional form does not take into account different energetic contributions that should be addressed. Right now we have considered adding correction terms regarding the strain and the conformational entropy loss of the ligand upon binding, as in eq. 2. ΔG = ΔGbe + ΔHstrain − TΔSconf (2)
