Automated Analysis of Conformational Flexibility in Small Molecules Using SAMPL7 Physical Properties Challenge Datasets from Alchemical Free Energy Simulations with MDPOW

Analysis of conformational flexibility is essential in identifying differences in solute-solvent and electrostatic interactions of small molecules. MDPOW (https://github.com/Becksteinlab/MDPOW), a python package, automates calculation of water-solvent partition coefficients from free energy perturbation (FEP) simulations, but subsequent analysis requiring user selection of each atom group poses a problem when sampling the entire conformational space, so there is a need to automate the process of obtaining relevant atom groups and their dihedral angles. Our extension of MDPOW automatically obtains dihedral atom groups from each molecule’s simulation ensemble and the kernel density estimation (KDE) of their dihedral angle frequencies, comparing across solvents and intermolecular interactions. Testing was done on large datasets from MDPOW generated for the SAMPL7 physical properties challenge, including simulation data that returned log P_ow estimates within a minimum precision level of 0.1 log units for AMBER/GAFF force fields. This development provides an automated workflow for sampling the whole conformational space of small molecules. Future adaptation for additional observables (bond length, hydrogen bonding) and implementation of Random Forest classification for measuring convergence will result in complete automation of physical properties analysis across libraries of small molecules, promising useful applications in screening for potential drug candidates.