Is it possible to run non-equilibrium FEP from pre-equilibrated end points?

[DreamCykl]

Hi,

I've been following the pmx tutorial found here. In this tutorial, the hybrid molecule with both A and B end states is created at the start of the protocol, and subjected to energy minimisation, equilibration and then transition stages.

I was wondering whether it is possible to create the hybrid molecule post-equilibration, and pre-transition. I.e., Run the EM and EQ protocol with a single molecule A, extract the frames, and then at every frame you combine molecule A with molecule B to create the hybrid molecule using PMX ligandHybrid. I guess you would have to do some sort of alignment procedure to ensure that B is aligned with the positions of A in the frame. You would also have to do some fiddling with the coordinate files to extract the original molecule and re-insert the hybrid molecule.

Do you have any ideas whether this would work, and do you have any advice for achieving this?

Thanks.

[vgapsys]

yes, you can do this.

You need to perform atom mapping and create topology only once. After running the simulations you can use ligandHybrid on the molecule at each frame. Note, you will need to create different topologies for stateA and stateB. Also, later when analysing the results, you will need to use --reverse flag to flip the sign for the reverse distribution

[DreamCykl]

Many thanks - I was struggling to get it to work until you pointed out the --reverse flag option.

I have a quick question about pmx ligandHybrid, which I'm running in conjunction with atomMapping to create the pairs.dat input file. Does ligandHybri align the two input coordinates when it generates the merged coordinates? If so, how does it achieve this?

[noahharrison64]

Hi @vgapsys

I have implemented a version of this. The correlations to experimental data for the non-hybrid eq endpoint FE calculations are the same as using hybrid eq endpoints, and the RMSEs are reasonably similar. However the convergences (BAR estimated Conv) tend to be a lot worse.

Do you have any suggestions for why this may be? I wonder if having to align lig_B to lig_A at each snapshot may be a contributing factor. Any thoughts/suggestions would be appreciated.

[vgapsys]

Could it depend on how gentle the equilibration of each snapshot is after creating a hybrid? I would suggest several intermediate steps: (i) freeze everything except dummies and minimize energy; (ii) run short (e.g. 20-50 ps) equilibrium simulation with the hybrid structure to equilibrate velocities; (iii) run the transition