As for my knowledge, binding energy can be obtained from the potential energy per atom.

There's a plugin in VMD, BFEE. You can obtain more information here: Article BFEE: A User-Friendly Graphical Interface Facilitating Absol...

Best,

Lorenzo

How many parallel MD simulations? How many MD snapshots? How many distinct conformational states? Have you perform also MD simulations of free ligand in solution?

Shaimaa A. Gouhar I think this link can help you. one of the comments contain a way to run MMGBSA using Ambertools on the dcd trajectory file from namd

https://www.researchgate.net/post/Can_I_use_MMPBSApy_script_on_dcd_trajectories_produced_by_NAMD_or_not

Thank you sir @Martin Klvana

I made MD simulation for 100 ns and have 500 frames, 8 conformational states, I did MD for ligand in solution using CHARMMi-GUI

Sir @Lorenzo Briganti

Thank you I will go through the link and try.

Sir @Yao Du

Thank you for your answer

500 frames might be too few, and the uncertainty in the predicted binding free energies too high. The uncertainty can be calculated as the difference between binding free energies calculated separately for the first and second half of the simulation; if it is higher than 2 kcal/mol, additional simulations are needed.