Simulating Large-Scale Conformational Changes of Proteins by Accelerating Collective Motions Obtained from Principal Component Analysis

Abstract

Enhanced sampling methods remain of continuing interest over the past decades because they are able to explore conformational space of proteins much more extensively than conventional molecular dynamics (MD) simulations. In this paper, we report a new sampling method that utilizes a few collective modes obtained from principal component analysis (PCA) to guide the MD simulations. Two multidomain proteins, bacteriophage T4 lysozyme and human vinculin, are studied to test the method. By updating the PCA modes with a proper frequency, our method can sample large-amplitude conformational changes of the proteins much more efficiently than standard MD. Since those PCA modes are calculated from structural ensembles generated by all-atom simulations, the method may overcome an inherent limitation called "tip effect" that would possibly appear in those sampling techniques based on coarse-grained elastic network models. The algorithm proposed here is potentially very useful in developing tools for flexible fitting of protein structures integrating cryo-electron microscope or small-angle X-ray scattering data.