There's no section on normal mode-based methods

[jamesmkrieger]

Several of these are reviewed in https://pubmed.ncbi.nlm.nih.gov/32622329/ and I think they deserve a brief mention in your review too. I'm not sure how they fit into your categories

Some of them involve switching between normal mode and MD calculations (e.g. CoMD and ClustENM). This allows the collective variables to change as new conformations have different normal modes.

CoMD is a transition method based on targeted MD but it uses several rounds of targeted MD with updated targets based on Monte Carlo sampling of normal modes with acceptance modified by approach to the target with a tuneable parameter that can be set to zero for pure exploration.

ClustENM directly applies random combinations of normal modes to the coordinates without any targeting in small steps with relaxation of resulting conformations by energy minimisation and its variant ClustENMD includes some small MD runs too.

Another method called NMMD (which is primarily developed for fitting to EM maps but could be used without this option; https://pubmed.ncbi.nlm.nih.gov/35150654/) adds the normal modes to the coordinates like ClustENM but does it simultaneously with the standard MD coordinate updates using two velocity Verlet schemes.

Another approach is to add the normal modes to the velocities, effectively raising the temperature of selected degrees of freedom. I guess you have a section of methods doing that, but I'm not sure where.

An early method called amplified collective motions (https://www.sciencedirect.com/science/article/pii/S0006349503750905) did this continuously with two weakly coupled water baths (Berendsen thermostats) for the velocity component projected onto the normal modes and the remainder of the velocity. They also kept updating the normal modes after a certain number of time steps e.g. 100.

MD with excited normal modes (MDeNM) also has the normal modes added to the velocities but they do this with a fixed set of normal modes and just update the velocities instantaneously every few ps and let the temperature relax back down. The simulations would have to be stopped and started again to update the normal modes as it's not included in the algorithm, but this can likely be done fairly easily in the new implementation MDexciteR. There is also now a related method called kinetically excited targeted MD (ketMD): https://pubmed.ncbi.nlm.nih.gov/36016719/.

Normal modes have also been used as collective variables in parallel tempering metadynamics (e.g. https://pubmed.ncbi.nlm.nih.gov/33351613/).

There is also one where normal modes are incorporated as biasing potentials in Hamiltonian replica exchange (https://pubs.acs.org/doi/full/10.1021/ct7002258 and https://onlinelibrary.wiley.com/doi/10.1002/prot.24695)

All these methods could also be used with PCA components and there are several papers doing that too