The biggest problem with DFT as a docking score (that is, a tool to predict free energy of binding) is that it will not be very helpful. By using it you will only determine the interaction energy for a particular docking pose (which will not necessarily be more accurate than that calculated by an MM force field). Hence, the (extremely important) entropic component of binding will be completely neglected, thus leading to even greater inaccuracies in affinity predictions than those that result from using fast empirical docking scores.

Dear Prof. Kireev,

I thank you for your time and consideration. I would like to ask an other question.

If I had a situation, in which I have two docking score without any significantly difference between them for two different binding, could I use any ab initio methods to determine which is the best by increasing my method accuracy?

I always thinks, geometry optimisation using ab initio could give more accurate results, since ab initio consider all electrons energies, molecular dynamic takes them as one particle.

Thank you again, Prof.

Sincerely yours

The property we are looking to predict is free energy of ligand-protein binding (dG). Ligand-protein interaction energy for a single pose is a very poor predictor of dG, irrespectively of the method that was used to calculate it.