Kinetic Ensemble Refinement

This repository contains the code of the KEnRef (Kinetic Ensemble Refinement) library.

KEnREf is developed as a stand-alone high-performance C++ library with interface(s) to interact with the major molecular modeling software packages. Currently, Gromacs only is supported.

Click here to go to Smith lab for more information.

Why KEnRef

KEnRef unique advantage relative to all previous methods, is that it accounts for the effects of angular motion, not only the interactomic distances .

Angular Motion	Distance Fluctuation

How it works

NOE Cross Relaxation Rate value $\sigma$ is used to derive the restraint energy from the simulation.

Is it reliable?

Average Structure Accuracy

RMSDr of the $r^{-6}$ averaged distances shows gradual decrease with higher force constants.

The higher the K:

• the lower the $RMSDr$,
• the greater the gap between each restrained subset and its related test subset, starting with a small gap in case of $K=10^8$.

Angular Dynamics

• Rigid restraints: KEnRef achieved high overall angular order with average $S2$ of 0.97 (matching start) and 0.95 (non-matching start) vs. average $S2$ of 0.85 for unrestrained simulations.
• Dynamic restraints: Target average $S2$ was close to the unrestrained simulations. All simulations started from the dynamic ensemble stay at order parameters ~ 0.85. KEnRef simulations converge faster than the unrestrained simulations, with the speed of convergence increasing with the $K$ force constant.
• Correlation function plateau values are sufficient to produce ensembles that match the overall angular motion of a structural ensemble used to produce that data.

Distance Dynamics

• Average standard deviation of interatomic distances [Average $\sigma(r)$], represents the overall amount of distance dynamics in the ensemble.
• Correlation coefficient between the restrained simulations and the reference, [$R(\sigma(r))$], indicates whether atomic fluctuations are located in the correct parts of the structure.

The higher the $K$,

• the closer and faster the simulations stick to, or go to the reference average values of 0.0 Å (rigid simulations) or ≈ 0.38 Å (dynamic simulations),
• the more correlated the simulations are to the reference.

Test sets monotonically improve with increasing $K$, although the gap between each restrained and its test set widens.

KEnRef Structure

KenRef can be installed on top of Eigen, which itself is a cross-platform framework with implementations for Windows, Linux, and Mac OS.

Requirements

1. C++17 enabled compiler
2. Eigen C++ library
3. A recent CMake
4. Gromacs is required to install KEnRef-Gromacs Bridge