Can you select subsets of the system potential energy as CVs?

[Billy Noonan]

Dear Plumed Users,

I know you can define the entire system potential energy as a CV. But can you select any component of the system and print its energy?

For example, if you wanted make the dihedral angle energy of a protein motif a CV, could you do this? What about the potential energy of any solute - could that be a CV? 

This would be very helpful for use as CVs in WT-MetaD variants, or with GAMD, or to couple with parallel tempering. Is anyone working on implementing this if it is not already implemented?

Kind regards,

Billy

[Billy Noonan]

Dear All,

I have contributed to the discussion here by adding a post at the end:
https://gromacs.bioexcel.eu/t/gaussian-accelerated-md-gamd/1143/8

Kind regards,

Billy

[Michele Invernizzi]

Hi Billy,

Unfortunately it's not easy to use just a component of the energy as CV, because it depends on the details of how the underlying MD engine works.
Probably the most promising approach to do this with GROMACS is to follow the steps of alchemical metadynamics https://github.com/shirtsgroup/alchemical_metadynamics_project and write an EXTRACV that gets what you need. It's something I've been wanting to try during my PhD, but never got the time to actually work on it.

Regarding implementing GAMD with PLUMED, you might be interested in the last section (1.4 - Bonus exercise) of this tutorial https://github.com/invemichele/masterclass-22-03/blob/master/notebooks/1-opes_expanded.ipynb

Best,

Michele

--
You received this message because you are subscribed to the Google Groups "PLUMED users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/b6c753cd-d94c-4bb9-9420-c08d8ef81f6fn%40googlegroups.com.

[Michele Invernizzi]

one last thing, another approach that I don't recommend, is what we did here https://journals.aps.org/prx/abstract/10.1103/PhysRevX.10.041034 for the water example, which is to simply write a CV that computes the part of the energy you need (in that case LJ interactions). It works, but performance is quite bad.

Michele

[Billy Noonan]

Hi Michele,

Thank you for your insight!

I am still struggling to understand the difficulties in integrating the use of partial elements of the potential energy function.

I gave dihedral term energies and solute energies as examples.

In terms of dihedral energy for connected atoms (bonded energy) - that is just a harmonic so why not have code that calculates that energy every time step and prints it with a frequency? I suppose could write that in Plumed using MATHEVAL and DIHEDRAL ATOMS taking the force constants from the topology - but this does not seem difficult to integrate?

With respect to energy groups overall - you can print the energy of any group if you put that group in your .mdp file. Couldn't Plumed work with whatever portion of the code which does that?

I have never looked at the GROMACS code but maybe that would help me understand the difficulty..

Thanks again for your links and papers.

Kind regards,
Billy

You received this message because you are subscribed to a topic in the Google Groups "PLUMED users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/plumed-users/WLCN5WXkpqg/unsubscribe.
To unsubscribe from this group and all its topics, send an email to plumed-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/CAGBhpifaq80BB250w-4iuP_OGWFodznTa%2BPfpo8mwJjt_%2BZbFA%40mail.gmail.com.

[Michele Invernizzi]

Hi Billy,

In terms of dihedral energy for connected atoms (bonded energy) - that is just a harmonic so why not have code that calculates that energy every time step and prints it with a frequency? I suppose could write that in Plumed using MATHEVAL and DIHEDRAL ATOMS taking the force constants from the topology - but this does not seem difficult to integrate?

Yes, definitely doable. Main drawbacks are that it is a boring process that has to be redone for each new system and that it can be quite slow to recalculate the energy term a second time with PLUMED. For the LJ term in our paper, we coded it in cpp because it was faster than using the PLUMED input file.

With respect to energy groups overall - you can print the energy of any group if you put that group in your .mdp file. Couldn't Plumed work with whatever portion of the code which does that?
I have never looked at the GROMACS code but maybe that would help me understand the difficulty..

You need access not only to the value, but also to the derivatives to use it for biasing (again, see EXTRACV and/or alchemical metad). It might be easy or very hard, I don't know.

Best,
Michele

To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/CAE9vE8cifsVFU1JP3D9VQ8ZYijrU3R6SrwoennxhpHR5mmspgw%40mail.gmail.com.