Motivation
Molecular docking is an efficient method to predict the conformations adopted by the ligand within the target binding site. Usually, standard docking protocol involves only one structure to represent the receptor, overlooking the changes in the binding pocket geometry induced by ligand binding. In our previous work, we observed that different conformations of the same target show different volume and shape of the internal cavities (Sessa et al., 2016). Different ligands may stabilize different receptor conformations with different internal cavities. Consequently, the crystallographic data represent the adaptation of a protein to a particular ligand. Cross-docking is a validation procedure consisting in docking a series of ligands into different conformation of the same receptor. Since the structures of the same receptor can be rather different, the cross-docking analyses are typically very poor. In these cases the internal cavity of the buried binding pocket does not have space enough to accommodate all ligands and this can radically affect the outcome and alter the cross-docking results. The changes of the cavity volume might explain the failure of traditional docking method and support the hypothesis that a single representative structure for the receptor is not enough. Keeping target proteins flexible during the docking has a high computational cost. To overcome this limit, our docking strategy is to represent receptor flexibility through an inexpensive method that generates a series of target structures. Starting from a known target structure, we used the molecular dynamics (MD) simulations to explore the conformational changes induced by ligand binding and to collect several snapshots of receptor structures to perform the cross-docking studies. To validate the accuracy of our flexible protocol in docking, we used a set of 10 crystallographic conformations of Androgen Receptor with the same target but with a different ligand. We performed two parallel experiments of docking, one with a rigid protein target and one considering flexible receptor structures. In addition, we compared the results for both experiments in the re-docking and in the cross-docking analysis.
Methods
Ten receptor structures complexed with a ligand were extracted from the X-ray structures in the PDB database (Berman et al., 2000). Several conformations for each receptor were selected from the molecular dynamics simulations (MD) at regular time intervals (each 500 ps). The MD simulations were performed with the software YASARA Structure 16.2.14 (Krieger & Vriend, 2014) using AMBER14 as force field. The molecular docking simulations were performed using VINA provided in the YASARA package.
"Abstract truncated at 3,000 characters - the full version is available in the pdf file"
Cross‐docking benchmark for automated pose and ranking prediction of ligand binding