Section: New Results

Orientational potential for small molecules

With Pablo Chacon and Karina Dos Santos Machado we have developed a novel statistical protein-ligand scoring function called KORP-PL. KORP-PL is based on the coarse grained backbone-only representation of protein that is very common in protein structure modeling. It is based on ideas implemented by Pablo Chacon in the preceding scoring function for protein quality assessment called KORP  [49], where each amino acid residue was characterized with a 3D oriented frame. We kept this representation for protein residues and computed the distances and angles between the oriented frames and points that describe ligand atoms. Using this data, we then derived the scoring function statistically in the same way, as it was done for protein-protein interactions in KORP. We also ran an optimization procedure to compute weights for each residue-atom interaction that would minimize the difference between predicted and experimental binding constants. We have assessed KORP-PL on several benchmarks  [47], [67], one of which was manually derived from the user-submitted data of the D3R Challenges 2  [35], 3  [36], and 4. On all of them it performed exceptionally in pose prediction despite being a coarse-grained scoring function. We are currently preparing the corresponding manuscript for submission.