SAMPL9 blind predictions using nonequilibrium alchemical approaches

TL;DR

This paper demonstrates the successful application of a nonequilibrium alchemical approach to predict host-guest binding free energies in the SAMPL9 challenge, showing improved performance over previous methods.

Contribution

It introduces a virtual double system single box approach combined with nonequilibrium simulations for accurate binding free energy predictions in host-guest systems.

Findings

Acceptable Pearson and Kendall coefficients indicating reliable predictions.

Better performance than previous SAMPL host-guest predictions.

Identified large mean signed error due to charge correction assumptions.

Abstract

We present our blind predictions for the Statistical Assessment of the Modeling of Proteins and Ligands (SAMPL), 9th challenge, focusing on binding of WP6 (carboxy-pillar[6]arene) with ammonium/diammonium cationic guests. Host-guest binding free energies have been calculated using the recently developed virtual double system single box approach, based on the enhanced sampling of the bound and unbound end-states followed by fast switching nonequilibrium alchemical simulations [M Macchiagodena, M Pagliai, M Karrenbrock, G Guarnieri, F Iannone, Procacci, J Chem Theory Comput, 16, 7260, 2020]. As far as Pearson and Kendall coefficients are concerned, performances were acceptable and, in general, better than those we submitted for calixarenes, cucurbituril-like open cavitand, and beta-cyclodextrines in previous SAMPL host-guest challenges, confirming the reliability of nonequilibrium approaches for absolute binding free energy calculations. In comparison with previous submissions, we found, nonetheless, a rather large mean signed error that we attribute to the way the finite charge correction was addressed through the assumption of a neutralizing background plasma.