Using Dimensionality Reduction to Systematically Expand Conformational Sampling of Intrinsically Disordered Peptides.

Abstract

One of the approaches to improve our ability to characterize biologically important processes and to map out an underlying free energy landscape is to direct MD simulations to explore molecular conformational phase space faster. Intrinsically disordered systems with shallow free energy landscapes of a huge number of metastable minima pose a particular challenge in this regard. Both characterization of the often ill-defined conformational states as well as the assessment of the degree of convergence of phase space exploration are problematic. We have used a multidimensional scaling-like embedding (sketch-map) to describe the energetically accessible regions of phase space for a peptide fragment of the intrinsically disordered protein α-synuclein. Using sketch-map coordinates from a short initial simulation, we guided additional MD simulations to efficiently expand sampling of the conformational space. The sketch-map projections are very well suited to detect rare but possibly functionally relevant events, metastable intermediates, and transition states in the vast amount of data.