From experimental, structural probability distributions to the theoretical causality analysis of molecular changes

Abstract

A brief overview of causality analysis (CA) methods applied to MD simulations data for model biomolec ular systems is presented. A CausalMD application for postprocessing of MD data was designed and implemented. MD simulations of two model systems, porphycene (ab initio MD) and HIV-1 protease (coarse-grained MD) were carried out and analyzed. Granger's causality methodology based on a Multivariate Autoregressive (MVAR) formalism, followed by the Directed Transfer Function (DTF) analysis was applied. A novel approach based on the descriptors of local structure was also presented and preliminary results were reported. Casuality analyses are required for a better understanding of biomolecular functioning mechanisms. In particular, such analyses can link physics-based structural dynamics with functions inferred from molecular evolution processes. Current limitations and future developments of the presented methodologies are indicated.