A High-Dimensional Lab-Frame Model of Markovian Dynamics in Alanine Dipeptide

Abstract

We present a high-dimensional coarse-grained model of alanine dipeptide that explicitly incorporates global translation and rotation alongside internal conformational dynamics. Using atomistic molecular dynamics trajectories, we construct a 10-bead Markovian Langevin model with a 24-dimensional internal potential of mean force and a full matrix of site-specific friction coefficients. The equilibrium landscape is represented by a Gaussian mixture model, while a rotational-entropy correction accounts for the coupling between internal conformation and global molecular motion. Friction coefficients are obtained using a generalized Einstein relation based on time-correlation functions. We show that the rotational-entropy correction is essential for obtaining a physically admissible friction matrix and significantly affects the inferred dynamics. Without fitting to coarse-grained dynamical observables, the resulting model reproduces a broad range of atomistic auto- and cross-correlation functions. The result is a compact and interpretable dynamical representation of the entire peptide at the chosen CG resolution, which explicitly preserves distinctions between energetic and dissipative coupling over a wide range of motions, including those between conformational fluctuations and global translational and rotational diffusion.