We present a general approach to transform between molecular potential functions during free energy calculations using a variance minimized linear basis approach. This approach splits the potential energy function into a sum of pairs of basis functions, which depend on coordinates, and `alchemical' switches, which depend only on the coupling variable. This approach for low variance molecular transformations is an alternative to to soft core potentials, which cannot be decomposed in such a manner. The first advantage of this approach is the calculation of the coupling parameter dependent terms is removed from inner loop force routines. This means that free energy-related calculations do not not have to be performed in the inner loops, significantly improving the ease of writing free energy code on highly parallelized architectures such as GPU's or Many Integrated Core (MIC) platforms. The second advantage is the flexibility in specifying basis functions and alchemical switches allows users to choose transformation pathways which maximize statistical efficiency. The third strength of the method is the ability to predict entirely in post-processing the thermodynamic properties along any alchemical path from a single simulation. This allows construction of optimized, near minimum variance alchemical switches without iterative simulations. We describe how to construct these basis functions in two classes of alchemical transformation, solvation and change of atomic identities. We apply this method to real molecular systems of different sizes, shapes, and charge distributions to find low variance paths connecting the end states. We also recommend an efficient general set of basis functions and alchemical switches for both classes of alchemical transformation.

Linear basis function approach to efficent alchemical free energy calculations

247th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.