RE: Targeted molecular dynamics domain selection question

From: Vermaas, Joshua (
Date: Fri Mar 01 2019 - 12:05:50 CST

Hi Aditya,

You are fine. Many of the other times when NAMD asks for an input PDB, it will let you specify which column to use. This just happens to be one of the exceptions where it defines the input style.

>From my own academic path, I've had it beaten into me that RMSD isn't always the best reaction coordinate. If it turns out your non-equilibrium work is low using RMSD, you are probably fine, but if you can come up with something cleverer, you might come up with a better/more realistic pathway. One example of this is in 10.1021/ct5002285. Figure 3 highlights different potential pathways, with the red pathway coming from a 1-step TMD type simulation. Its non-equilibrium work is highest against the others, and totally misses what should be an intermediate state in the "real" mechanism.


On 2019-03-01 08:09:13-07:00 Aditya Ranganathan wrote:

Thanks Joshua. I feel stupid for missing that detail. Set the occupancies now and not the beta.
With regards to using TMD and not colvars, I`m a little conflicted. I`m working with a situation where I have the crystal structures of the initial (open state of an enzyme) and target (closed state of an enzyme) states available. I think it may not be a bad idea to use RMSD in such a case. That is why I chose TMD. I would appreciate your comments on the same. I might do a similar calculation along other order parameters using colvars and see how the landscape looks in both the cases.

On Thu, Feb 28, 2019 at 7:52 PM Vermaas, Joshua <<>> wrote:
Hi Aditya,

>From the NAMD user guide describing how the TMD file is used:

TMDFile [$ <$] File for TMD information [$ >$]
Acceptable Values: Path to PDB file
Description: Biased atoms are those whose occupancy (O) is nonzero in the TMD PDB file. Fitted atoms are those whose altloc field is not ` ' or `0', if present, otherwise all biased atoms are fitted. The file must contain no more atoms than the structure file and those atoms present must have the exact same index as the structure file (i.e., the file may contain a truncated atom selection ``index [$ < N$] '' but not an arbitrary selection). The coordinates for the target structure are also taken from the targeted atoms in this file. Non-targeted atoms are ignored. The beta column of targetted atoms is used to designate non-overlapping constraint domains. Forces will be calculated for atoms within a domain separately from atoms of other domains.

You specified based on the beta column, not the occupancy. Is there a particular reason to use TMD rather than colvars? Colvars lets you do transitions between states with lower non-equilibrium work if you choose appropriate collective variables, while TMD as implemented I think can only use RMSD, which is not always the best collective variable to use.


On 2019-02-28 16:26:10-07:00<> wrote:

Dear All,
I had a syntactical question regarding targeted molecular dynamics simulations performed using NAMD. I`m working on a TMD simulation where I want to bias my protein from state A to state B. The system has 58500 atoms, of which 3314 atoms belong to the protein. I label the beta column of the protein atoms in both the input pdb file (stateA.pdb) and the TMD pdb file (target structure) with 1. The beta for all non-protein atoms are set to 0 in both these pdb files. I only want to apply the harmonic restraints to the protein alone and not any other atom in the system. The output file for this trajectory reads as follows:
"Info: TMD domain 0 has 55159 atoms 55159 fitted 55159 biased
Info: TMD domain 1 has 3341 atoms 3341 fitted 3341 biased
Info: 58500 TMD ATOMS
TCL: Running for 30000000 steps
TMD 0 Domain: 0 58.2916 58.2916
TMD 0 Domain: 1 58.2916 6.8374
Since the output reads 58500 TMD atoms, I want to clarify if the TMD is being performed on both the solvent and the protein here? Some of the literature on the semantics is a little unclear.

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:33 CST