Use the modeled complexes as the initial structures for parallel molecular dynamics simulations: generate structural ensembles of the complexes and calculate the protein-ligand binding free energies, e.g. using linear interaction energy (LIE) methods.

---

https://www.ncbi.nlm.nih.gov/pubmed/9570087

• Thank you Martin for your valuable advice. As before mentioned i still havent done any MD for the complex. Only for the cleaned crystal structure. Docking procedure was done on 6 of MD frames i obtained via intial MD equilibration for protein. Taking average binding energy isnt the best move as i thought. Any ither ideas?

With the data that you have you could run a PCA or other structural analysis and get the structural probability of the 6 frames that you chose and average the 6 frames weighted by there normalized probabilities (should be in total 100 %). You should keep in mind that unless you covered at least around 90 % of your structural space this value can have a large error. Additionally, energies obtained by docking alone can only be compared relative to each other and should never be seen as absolute values. If you want a more accurate binding energy use the approach proposed by Martin Klvana or do a FEP calculation (very time consuming).

If you you don't expect much entropy of either the protein or the ligand induced to the binding, running a single MD of the bound complex and using theMMGBSA method on it might be a valuable method to obtain well ranked binding free energies, including an approximation for the solvation free energy, but as already mentioned  lacking entropic contributions due to the unbinding.