Challenges in Reproducing ddG Estimation Results for 1STN

pmx
21

Hi Marc,

It’s good to know that you could figure out the issue.
It was indeed a good learning exercise.

All the best for the 1stn mutations.

–
Sudarshan

22

hello,

I was putting this on a break but now i continue with my alchemical transformation simulations. While going through the tutorial and your mdp files, I noticed that the MDP options do not include the generation of velocities, and there is no simulated annealing applied to ramp up the temperature.

Is this approach intentional? I am curious to understand the rationale behind not generating velocities in the simulations, as it seems unusual compared to typical molecular dynamics simulations. I would appreciate any insights you can provide on this aspect of the protocol.

Thank you for your time and consideration.

kind regards,
Marc

23

Hi Marc,

I also think velocity generation and simulation annealing are always the better practices. I would suggest you stick to it.

I don’t think there is any rationale behind not generating velocities. However, I think the results would be independent of this, as the system goes through a good equilibration phase.

Best,
Sudarshan

24

Velocity generation might perturb the system which you may not want. Starting the simulation without the velocities and allowing them to equilibrate gives a smoother equilibration (can couple it with simulated annealing). I would also recommend using sd integrator for free energy calculations

1 Like
25

Hi Vytas,
If we do not generate velocities for independent runs, how are we to get independent simulations for different replicas or is there something I am missing?

Would this also work for equilibrium simulations to start without velocities?

26

Even when starting without the assigned velocities, the simulations will quickly diverge (you can convince yourself by checking velocities and coordinates of two simulations already after several steps).

This can be used in equilibrium simulations. For transitions, velocities should be taken from the equilibrated snapshots (extract the snapshots from a trajectory with coordinates and velocities).

1 Like
27

What I meant was using equilibrium free energy calculations

28

This would work for equilibrium FEP as well, just need to discard the first part of trajectory for equilibration.

1 Like
29

Hi @mstieffe,

I wonder if you were finally able to reproduce the ddG data for 1STN.

I did a benchmark recently using gmx2022.6 and the pmx develop branch for both SNase and Barnase, and the metrics reported in the paper are well-reproduced.

Please find the attached PDF.
protein_mut_reproduction.pdf (7.0 MB)

Thanks,
Sudarshan