Is it possible to calculate absolute binding free energy with PyAutoFEP?

[Fairly]

Hi there, thank you for maintaining this awesome repo! I'm wondering whether it is possible to calculate absolute binding free energy with PyAutoFEP, either with the current code base or some modifications on it.

[luancarvalhomartins]

(I moved the issue to a discussion, as I believe it is more suitable here. Feel free to open a new one if you think otherwise.)

In principle, using the regular code, no. Because of the way dual-topologies are generated, a molecule must be perturbed into another, so the free energy will be relative.

However, I believe prepare_dual_topology.py could be used to perturb a molecule in a version of itself with all non-bonded interactions turned off. Therefore, it might be possible to use it to prepare the systems for running the legs named ΔG prot elec+vdw and ΔG solv elec+vdw in Aldeghi 2016 (Fig 2 below). But here comes the catch: the current code won't allow to perturb both Coulomb and VdW parameters of atoms at once. I am currently working on a updated version that - for other reasons - will allow such.

(Fig 2 from Chem. Sci., 2016,7, 207-218)

So, the complete answer is: not currently, it will likely be possible in the near future with some tweaking and some (minor?) coding. Furthermore, I think that this would be a quite useful addition to PyAutoFEP and a somewhat low hanging fruit - after the new code allowing VdW+Coulomb perturbation of constant region is deployed. I would be interested in adding it, specially in case there are people interesting in helping code and test.

[luancarvalhomartins]

Sorry, I should have unpacked this point. Currently, the terms in the lambda.ini are only applyed to the A and B regions of the dual-topology. In case one is perturbing a molecule in a version of itself, all atoms will match so that the entire molecule would be the "constant" region (grey atoms in the fig below). For this atoms, the current version only allows perturbing the Coulomb terms. That's because PyAutoFEP expect that the atomtypes are the same. However, for charged perturbations, pertubring both VdW and Coul terms at once is required to morph an water into an ion. So, when the charged perturbations code makes it in to the tree, allowing prepare_dual_topology.py to perturb a molecule should be simple.

PS: in case you are asking whether there should be another way to passing a lambda scheme to prepare_dual_topology.py, there is. See documentation for the lambda_input option.

[Fairly]

Oh, got it. Thank you so much for making things clear. Things seem complicated, but please let me know if any thing I could help you with.

[luancarvalhomartins]

I would love your (and others) help in implementing and/or testing this. I'm opening a discussion about absolute FE implementation after the code for charged perturbations makes it into the master branch.

[spadavec]

hello @luancarvalhomartins -- just curious, is this repo now in support of ABFE calculations?

[luancarvalhomartins]

Hi, @spadavec. Thanks for asking.

No, unfortunately, not yet. The code to allow perturbing VdW and lambda that I mentioned above is ready but not yet committed. I plan to do that together with the charged-perturbation code, which I wrote but was not able to test extensively yet.

After that, it could be possible to implement ABFE as in Chem. Sci., 2016,7, 207-218 with a moderate amount of coding.

As I mentioned above, it would be awesome if other were willing to help in coding and/or testing it :)