Umbrella Sampling MD Simulations for Retention Prediction\nin Peptide Reversed-phase Liquid Chromatography

Reversed-phase\nliquid chromatography (RPLC) is an essential tool\nfor separating complex mixtures such as proteolytic digests in bottom-up\nproteomics. There is growing interest in methods that can predict\nthe RPLC retention behavior of peptides and other analytes. Already,\nexisting algorithms provide excellent performance based on empirical\nrules or large sets of RPLC training data. Here we explored a new\ntype of retention prediction strategy that relies on first-principles\nmodeling of peptide interactions with a C18 stationary phase. We recently\ndemonstrated that molecular dynamics (MD) simulations can provide\natomistic insights into the behavior of peptides under RPLC conditions\n(<i>Anal</i>. <i>Chem</i>. <b>2023</b>,\n95, 3892). However, the current work found that it is problematic\nto use conventional MD data for retention prediction, evident from\na poor correlation between experimental retention times and MD-generated\n“fraction bound” values. We thus turned to umbrella\nsampling MD, a complementary technique that has previously been applied\nto probe noncovalent contacts in other types of systems. By restraining\nthe peptide dynamic motions at various positions inside a C18-lined\npore, we determined the free energy of the system as a function of\npeptide-stationary phase distance. Δ<i>G</i><sub>binding</sub> values determined in this way under various mobile phase conditions\nwere linearly correlated with experimental retention times of tryptic\ntest peptides. This work opens retention prediction avenues for novel\ntypes of stationary and mobile phases, and for peptides (or other\nanalytes) having arbitrary chemical properties, without the need for\nRPLC reference data. Umbrella sampling can be used as a stand-alone\ntool, or it may serve to enhance existing retention prediction algorithms.