Algorithmic Enhancements to the Solvaware Package for the Analysis of Hydration

Key Personnel

PI/Co-I: Dr David Huggins - University of Cambridge

Technical: Dr. Arno Proeme - EPCC

Relevant Documents

eCSE Technical Report: Algorithmic Enhancements to the Solvaware Package for the Analysis of Hydration

Project summary

The Solvaware package is a workflow that runs and analyses molecular dynamics (MD) trajectories to estimate hydration free energies by computing the contribution of different subvolumes around a solute. The goal of this project was to make the Solvaware package available on ARCHER and develop it by improving the efficiency of key algorithms in the workflow and increasing the functionality.

The accurate calculation of hydration free energies is a vital goal in computational modelling of biological and engineered aqueous systems. Water is the most important molecule in biology, as the interaction between biomolecules must compete with the interactions of the individual molecules with water. Thus water is directly involved in every single intermolecular interaction in an aqueous environment. Unfortunately, the complexity of intermolecular interactions and the vast number of degrees of freedom in even the simplest systems commonly confound computational approaches. In particular, the calculation of entropy remains a challenging and computationally demanding problem.

This project focuses on the statistical mechanical method of inhomogeneous fluid solvation theory (IFST) to determine the enthalpy and entropy of solvation. IFST can be used to analyse molecular dynamics (MD) simulations in explicit solvent and derive the contribution of specific spatial subvolumes to the solvation free energy.

IFST is unique in combining the accuracy of a rigorous treatment of explicit water with the spatial decomposition of the results, which leads to understanding and drives the generation of ideas. In addition, useful predictions can be derived from a single simulation. IFST analysis is performed by calculating the mean enthalpy and entropy in different subvolumes using snapshots from the MD trajectory (see below).

At the project outset, the Solvaware software package was only available on the Cambridge University High Performance Computing Service (HPCS). During the project, the software was rewritten using OpenMP and has now been made available as a software package on ARCHER.

The major improvement to the code was to the efficiency of the k-nearest neighbours (KNN) algorithm for entropy estimation, which resulted in a 25,000-fold speed-up. In addition, improvements to the MD protocol led to ~30% efficiency gain for simulations of typical systems. Given the improvements to the IFST code, this now constitutes a significant part of the total time spent obtaining results using Solvaware and makes previously unfeasible calculations now feasible.

Summary of the software

The enhanced Solvaware package for the analysis of hydration is available to use on ARCHER as a centrally installed package, loaded using the "solvaware" module. Access to the software is arranged via application to the ARCHER helpdesk, who will confirm registration of users with the PI.