AMBER takes in to account the Simulation Process i.e it takes the snapshots generated from the trajectory file and then calculates free energy of binding.

whereas

Prime MMGBSA estimates the relative binding affinity of Protein-Ligand without taking in to account of any simulation process.It is more of a general Calculation.

Both of them differ from each other. Since, Simulations are closer to realistic setup, AMBER MMGBSA are more accurate i would say.

References:

http://ringo.ams.sunysb.edu/index.php/2014_AMBER_tutorial_with_HIV_Protease#MM-GBSA_Energy_Calculation

https://www.schrodinger.com/kb/1647

Hi Vishnu,

For AMBER MM-GBSA, you should switch off the NMODE entropy calculation.

If your calculating binding free energy of simple protein-ligand complex, prime module use VSGB solvation model and calculation performed on the system are more generalized. You can use thermal_mmgbsa.py, python script provided by Schrodinger to calculate the binding free energy of simulated system for each trajectory using prime module.

In AMBER, interaction energy values are being averaged over MD simulation snapshots; in Maestro, over deterministically sampled ligand poses. It sounds like AMBER's way is more rigorous and, hence, should result in more accurate affinity predictions. However, one should keep in mind that, given popular simulation times ranging from 1 to 5 ns, the configuration space of ligand-protein interactions is, in general, severely undersampled. Hence, I would not give any a priory advantage to either approach. In the long run, I would expect them to perform equally poorly.