Challenges in Reproducing ddG Estimation Results for 1STN

Thank you very much for providing the mdp files. However, I have some questions regarding the velocity equilibration before the transition. In my current setup, I perform 20 ns equilibration runs for both the WT and MUT states. I then extract 100 equidistant frames from each of these runs and carry out a 50 ps transition for WT->MUT and MUT->WT, respectively. Is it necessary to perform velocity equilibration before the transitions in this setup?

I used your mdp files (for equilibration and transition) with GROMACS 2023.5 and the master branch of PMX, deploying the amber99sb-star-ildn-mut forcefield. I completed four runs each for F2A and I2A, obtaining -37.05±0.24 kJ/mol and -12.51±0.26 kJ/mol, respectively. The webserver reports -31.95±0.39 kJ/mol and -9.64±0.33 kJ/mol. Although the deviations are not overly large, they are still significant and do not appear to be outliers, as I consistently obtained these values over multiple separate runs. Interestingly, for the other two tripeptides I tested (V2A and L2A), I was able to perfectly reproduce the results from the webserver.