Gneimosim: Multiscale Internal Coordinates Molecular Dynamics for Proteins

Abstract

The Internal Coordinate Molecular Dynamics (ICMD) method is an attractive MD method that allows the study of the dynamics of proteins and other polymer in internal coordinates such as bonds, angles and torsions. The ICMD method is a convenient platform to coarsen the dynamics model at hierarchical levels by simply making the internal coordinates rigid and excluding them from the dynamics model, while still using accurate all-atom forcefields. This feature makes ICMD highly suitable for multi-scale simulations of protein macromolecular complexes.The coupled nature of the ICMD coordinates and the high complexity of ICMD equations required sophisticated mathematical techniques, that remained a bottleneck in its widespread adoption. Another long-standing bottleneck has been the increased rigidity in constrained ICMD methods, distorting the potential energy profiles and affecting the probability density functions of states and barrier-crossing rates. We have made significant advances in rigorously addressing these challenges through the Generalized Newton-Euler Inverse Mass Operator (GNEIMO) method, by implementing a fast and efficient algorithm to calculate the Fixman correction potential for branched molecules. Additionally, we have developed the Hybrid ICMD method that allows the user to free up the desired bond angle degrees of freedom, removing the distortions in the potential energy surface and yielding more accurate thermodynamic properties.Thus the GNEIMO ICMD method now provides an internal coordinates MD method with complete range of granularity ranging from all all-atom to coarse grain level of dynamics. We have demonstrated with examples that these methods together successfully remove the distortions in the potential energy landscapes observed in constrained ICMD simulations. We have used this integrated platform for the refinement of low resolution X-ray crystallographic data as well as Cryo-EM density to fit structural protein models.