CHARMM Development Project Forums User Discussion & Questions Molecular Dynamics PERT vs TSM in conjunctiion with OpenMM

Hi all,

I have a fairly simple soluble protein system (a kinase) with some of closely related drug-like ligands (~300Da) I'd like to do some relative alchemical-style free energy calculations on to qualitatively rank them based on some ITC data we have, predict new ligands, etc. So I'd like to decide on which module is the most likely to get me there in CHARMM.

My own sense is that TSM is perhaps the older set of modules and that PERT is more flexible (e.g. it has WHAM as an option but TSM doesn't?). The examples in the docs for PERT seem simple enough but then, of course, one still finds relatively modern tutorials for TSM (e.g. https://busathlab.github.io/mdlab/lab8.html). TSM seems to perhaps have the simpler setup. An additional question is whether anyone has tried these with the OpenMM interface and whether there have been any problems with that.

Are my assumptions wrong? Are there clear reasons why one would choose to use TSM over PERT or vice versa? I did a search on these methods and rmv has certainly commented on the fact that the different modules exist but can't find specific guidance on which might be more appropriate and under which circumstances this might arise. Perhaps there's a reason for that so I would appreciate any advice. Thanks guys.

Read openmm.doc to see what is available, and also check the testcases (test/c*test/omm*.inp). BLOCK is an option.

Thanks Lennart. I did see those but I see no mention in either of them of PERT or TSM. BLOCK has some limitations in the OMM implementation too. Of course, I don't expect you guys to explain OMM functions but I don't know whether absence of evidence equals evidence of absence in this case. Does this mean they are explicitly not supported? Is it simply unnecessary in OMM because running FEP/TI is fast enough with the CPU implementation perhaps?

A shame if it's not supported. I find BLOCK quite opaque. Stefan Boresch is a lovely dude and an utter wealth of information but the amount of options, tips, tricks, etc. in the CHARMM doc is pretty overwhelming. I've Googled and come up empty so if you're aware of something like an annotated MWE, I'd love to hear about it.

Thanks again.

You may be better off using the Python interface to OpenMM, which has a number of features not supported in the CHARMM implementation. There are provisions for FEP sampling.

In this case you can only use features that are explicitly stated to work (if in doubt test it yourself!). GPU is much faster than CPU.
I find "MWE" at least as opaque as the most obscure parts of the CHARMM documentation.
Yes, BLOCK has some limitations - but what exactly do you want to do?

Good to know - thank you.


heh, fair. MWE = Minimum Working Example. Comp sci jargon, sorry about that.


Nothing special, I'd say. Relative free energies of a series of drug-like ligands where we have some affinity data and crystal structures for the first few in the series. So, I'd like to see the changes in free energy for some minor extensions to these ligands in classic med chem fashion (e.g. change methyl to trifluoro, change ring -O- to -NH-, etc.). As they say, computational chemists have two jobs, convince experimentalists to do an experiment or convince them not to. So I'd like some support for an idea I have for what makes ligands bind in this particular protein and also confirm my idea for which part of the pocket to avoid with certain species, specifically to avoid synthetic dead-ends.

And thanks Rick, that's a good tip I'll look into.