In Deep Boosted Molecular Dynamics (DBMD), probabilistic Bayesian neural network models are implemented to construct boost potentials that exhibit Gaussian distribution with minimized anharmonicity, thereby allowing for accurate energetic reweighting and enhanced sampling of molecular simulations. DBMD has been demonstrated on a wide range of model systems, including alanine dipeptide and the fast-folding protein and RNA structures. Based on Deep Learning neural network, DBMD provides a powerful and generally applicable approach to boosting biomolecular simulations.
Gaussian Accelerated Molecular Dynamics (GaMD) is a biomolecular enhanced sampling method that works by adding a harmonic boost potential to smoothen the system potential energy surface. By constructing a boost potential that follows Gaussian distribution, accurate reweighting of the GaMD simulations is achieved using cumulant expansion to the second order. GaMD has been demonstrated on three biomolecular model systems: alanine dipeptide, chignolin folding and ligand binding to the T4-lysozyme. Without the need to set predefined reaction coordinates, GaMD enables unconstrained enhanced sampling of these biomolecules. Furthermore, the free energy profiles obtained from reweighting of the GaMD simulations allow us to identify distinct low energy states of the biomolecules and characterize the protein folding and ligand binding pathways quantitatively.
GLOW integrates Gaussian accelerated molecular dynamics (GaMD) and Deep Learning (DL) for free energy profiling of biomolecules. First, all-atom GaMD enhanced sampling simulations are performed on biomolecules of interest. Structural contact maps are then calculated from GaMD simulation frames and transformed into images for building DL models using convolutional neural network (CNN). Important structural contacts can be determined from DL models of saliency (attention) maps of the residue contact gradients, which allow for the identification of system reaction coordinates . Finally, free energy profiles of these reaction coordinates are calculated through energetic reweighting of GaMD simulations.
A toolkit of python scripts “PyReweighting” is provided for reweighting of aMD simulations. PyReweighting implements a list of commonly used reweighting methods, including (1) exponential average that reweights trajectory frames by the Boltzmann factor of the boost potential and then calculates the ensemble average for each bin, (2) Maclaurin series expansion that approximates the exponential Boltzmann factor, and (3) cumulant expansion that expresses the reweighting factor as summation of boost potential cumulants.
MD/OPX is designed to simulate large bionanosystems over long time periods by using short replica MD run(s) to extrapolate the structural order parameters (OPs) of the system over large time intervals and thus advance the system over long time. The present implementation of MD/OPX is based on NAMD using its Tcl scripting interface and requires NAMD for running. A Fortran code is used to read the output structure of a short dt NAMD run, calculate the resultant OPs, extrapolate the OPs for dT, generate an atomic configuration at t+dT with the extrapolated OPs and put the result all-atom structure back into NAMD to start the next (dt, dT) cycle.
Following is a list of TCL scripts that can be executed with VMD (http://www.ks.uiuc.edu/Research/vmd/) using “vmd -dispdev text -e *.tcl”. They can be used to prepare systems for NAMD simulations and analyze simulation output trajectories, notably on viral capsids.