Machine learning analysis of SH-dipeptide dynamics from atomistic molecular dynamics simulations.

Abstract

Serine-Histidine (SH-dipeptide) is a minimal peptide system with both hydrolase and peptide synthesis activity, and has been argued to be a potential candidate proto-enzyme in exobiological and/or origin-of-life settings. Here, we use a combination of machine learning and atomistic molecular dynamics to characterize the conformational landscape and dynamics of SH-dipeptide at ambient temperature, as a function of pH. We also compare the behavior of SH-dipeptide with Serine-Histidine-Aspartic Acid (SHD), a small peptide that emulates the most common serine hydrolase catalytic triad. We focus on identifying low-dimensional degrees of freedom that optimally impute the trajectory of the peptide as a whole, providing a framework that can be used for further studies of this system.