Trying to recompute interaction energy entropy but my nuber is completely off

[Hideya Koizumi]

Dear Sir,

I have extracted the interaction energy from result.csv.  I made sure in my mmpbsa.in that I uses all frames.  Then, I use these energies into the formula in published 2016 JACS paper.  However, the interaction energy entropic contribution -TDS =14kcal/mol whereas program result in 50kcal/mol.  (I can see this in result.dat or gmx_MMPBSA_ana)Am I suppose to extract the Coulomb and van de Waals directly from GROMACS to get this? or I can simply use DU in csv file?  I also tried the Var/2kT but they are way off from 50 kcal/mol.     Thank you very much for your guidance.   

[mariosergi...@gmail.com]

Did you consider the interval used to compute IE? Normally, internal energy is computed for the entire trajectory, but IE is computed only for the interval in which it is already stable. This can cause considerable variations in the final values. Make sure that the same interval is used in your calculation and in gmx_MMPBSA. In gmx_MMPBSA, the default is the end 25% of the trajectory. For example, if you're computing a trajectory of 1000 frames, internal energy would be calculated for all 1000 frames, but IE would be calculated for the final 250 frames.

[mariosergi...@gmail.com]

Please, attach the mmxsa file to check it

El sábado, 25 de octubre de 2025 a las 10:28:27 UTC-5, Hideya Koizumi escribió:

Thank you very much for your response. Yes, I initially suspected that might be the cause. To test the output without the segmentation effect, I used all 400 points. The output displayed the following message:
“Calculations performed using 400 complex frames. Interaction Entropy calculations performed using last 400 frames.”

For this test, I changed ie_segment from 25 to 100 to examine how it affected the output. I also verified the results using both Var/2kT, kT ln⟨e^βΔE⟩, and by computing ΔG to obtain –TΔS. However, all of these yielded around 14 kcal/mol (consistent with the gas-phase values reported in the article), compared to about 50 kcal/mol from gmx_MMPBSA.

Interestingly, ΔH is the only term that matches exactly between my results and MMPBSA, suggesting that the data used for the IE entropy calculation itself is correct. I have checked many possible sources of discrepancy but have not yet found the reason. I’m attaching my data here in case it helps clarify the issue.

[Hideya Koizumi]

Thank you very much for your response. Yes, I initially suspected that might be the cause. To test the output without the segmentation effect, I used all 400 points. The output displayed the following message:
“Calculations performed using 400 complex frames. Interaction Entropy calculations performed using last 400 frames.”

For this test, I changed ie_segment from 25 to 100 to examine how it affected the output. I also verified the results using both Var/2kT, kT ln⟨e^βΔE⟩, and by computing ΔG to obtain –TΔS. However, all of these yielded around 14 kcal/mol (consistent with the gas-phase calculation scheme), compared to about 50 kcal/mol from gmx_MMPBSA.

Interestingly, ΔH is the only term that matches exactly between my results and MMPBSA, suggesting that the data used for the IE entropy calculation itself is correct. I have checked many possible sources of discrepancy but have not yet found the reason. I’m attaching my data here in case it helps clarify the issue.  (This is a repost of previous message which I accidentally erased)

[mariosergi...@gmail.com]

The *.mmxsa file is a binary file that has all the serialized result data, not the mmpbsa.in file.