High differences between the computed binding energy with g_mmpbsa and experimental binding energy

[alireza shokrollahi]

Dear developers of g_mmpbsa,

I am working with g_mmpbsa to compute the binding energy of CYP17A1 and abiraterone. The CYP17A1 is an enzyme with heme as a cofactor. This heme plays a crucial role in substrate binding. Abiraterone is an inhibitor of CYP17A1. I performed three MD simulations for this complex system over 100ns. Then, I considered the last 50 ns with 0.5 ns intervals to compute the binding energy. For this purpose, I considered protein + heme as the protein group and abiraterone as the ligand. I got these results:

Repeat1:

Energy,Average,Standard-Deviation,
vDW,-184.452,8.267,
Electrostatic,-8.135,5.100,
Polar-solvation,74.786,13.205,
Non-polar-solvation,-18.610,1.180,
Total,-136.411, 11.551,

Repeat2:

Energy,Average,Standard-Deviation,
vDW,-192.830,8.699,
Electrostatic,-5.675,4.838,
Polar-solvation,65.721,14.531,
Non-polar-solvation,-19.179,1.102,
Total,-151.963, 12.409,

Repeat3:

Energy,Average,Standard-Deviation,
vDW,-205.854,9.656,
Electrostatic,-10.579,3.346,
Polar-solvation,82.965,8.047,
Non-polar-solvation,-19.975,1.100,
Total,-153.443, 10.381,

Based on the experimental data, the binding energy for this complex is approximately -53.6 kJ/mol. Why are there high differences between the computed binding energy with g_mmpbsa and the experimental binding energy? 

Kind regards,

Alireza Shokrollahi.