(357b) Free Energy Calculation with Non-Uniform Windows for Improved Computational Efficiency: A Proof of Concept

Performing free energy calculation using computational methods is a costly process. It involves selecting multiple parameters for the simulation using which the calculation is made. Umbrella sampling (US) is one of the go-to free energy calculation methods to achieve enhanced sampling. It is followed by techniques like WHAM (Weighted Histogram Analysis Method) to determine the histogram weights for the calculation. The resultant free energy is expressed in terms of the Potential of Mean Force (PMF). To perform the US and the WHAM calculations, the user selects the reactions coordinate (x), no. of windows, force constant (K) and the sampling time (ns) etc. These parameters are coupled and their relationship is not linear for most systems, especially for complex systems like polymers and biomolecules. Polymers are usually difficult to simulate as the model must be preprocessed before the user can derive useful information from it, for instance, relaxing the structure. Currently, the parameters are selected based on literature or trial and error. This is a tedious and computationally tolling routine. We are proposing an algorithm that selects the parameters and iteratively optimizes them using the Nelder-Mead algorithm, which is a simplex optimization method. Our algorithm, COWBOE (Construction Of Windows Based On Energy) performs the task starting from a guide PMF obtained from a fraction of the sampling time required for trial and error. The module has been made into a python package for light and easy installation.

Research Interests: Computer aided drug design (CADD), Molecular simulation, Free energy calculations, Biomolecular simulation, Polymer engineering, ligand docking.