Multiple alchemical regions

[amrhamedp]

Hello,
I am trying to make set of moves in which different regions (atom indices) should be perturbed. I am not sure how can this be handled in the current infrastructure. If I set up simulations factory , the alchemical system is set once according to the set of atoms in the first move (index 0) and thus becomes constant.

I will appreciate your help in this

[davidlmobley]

cc @nathanmlim

My understanding is that openmmtools does not yet support multiple alchemical regions; I believe someone in David Huggins' group is working on that. (Is that you?) If so we'd love to talk, as we very much want to do this and need such support before we can do so.

[davidlmobley]

(If it's not you, we should talk to Huggins and then get back to you.)

[amrhamedp]

Thanks Dr Mobley for replying. No I am not affiliated with David Huggins. I was just trying to implement a move that change randomly the atoms to be perturbed every MC trial. I looked at the side chain movement in blues implementation and it looks like all the rotatable atoms are subjected to alchemy while the movement do a random choice on one of them each time. Do you think that this will not impact the acceptance probability ?

[davidlmobley]

@amrhamedp - thanks for clarifying.

In terms of current moves with side chains, my understanding (but @nathanmlim can also comment; the main side-chain person is out on maternity leave right now) is that it should be selecting the atoms of a specific sidechain as the alchemical region, and subjecting those atoms (the set) to alchemical changes while it moves them. So that should be consistent, and I believe you SHOULD be able to do this for multiple different side chains as long as you do them one at a time.

What is NOT presently supported via openmmtools (which we use) if I understand correctly is to have multiple regions treated alchemically simultaneously.

@nathanmlim @sgill2 is that correct?

[davidlmobley]

@sukanyasasmal might also be able to comment.

@amrhamedp also it might help if you could elaborate on this:

I was just trying to implement a move that change randomly the atoms to be perturbed every MC trial.

What exactly are you trying to do? Are you trying to change which atoms within a given sidechain are perturbed? Or to change which sidechains are being rotated? Or...?

[amrhamedp]

The movement which i want to do involves separate molecules in which one molecule is picked up randomly and perturbed. It is basically different from the article example in which you have single ligand that is under investigation. Also it is not the same ligand with different conformations as in pose darting so basically the molecules which are under investigation are different.

[sukanyasasmal]

Are you working with protein-ligand system? How many ligands does your system have? What kind of moves are being attempted - random rotational ligand moves or any custom move?