Holo-like and Druggable Protein Conformations from Enhanced Sampling of Binding Pocket Volume and Shape

Abstract

Understanding molecular recognition of small molecules by proteins in atomistic detail is key for drug design. Molecular docking is a widely used computational method to mimic ligand−protein association in silico. However, predicting conformational changes occurring in proteins upon ligand binding is still a major challenge. Ensemble docking approaches address this issue by considering a set of different conformations of the protein obtained either experimentally or from computer simulations, e.g., molecular dynamics. However, holo structures prone to host (the correct) ligands are generally poorly sampled by standard molecular dynamics simulations of the apo protein. In order to address this limitation, we introduce a computational approach based on metadynamics simulations called ensemble docking with enhanced sampling of pocket shape (EDES) that allows holo-like conformations of proteins to be generated by exploiting only their apo structures. This is achieved by defining a set of collective variables that effectively sample different shapes of the binding site, ultimately mimicking the steric effect due to the ligand. We assessed the method on three challenging proteins undergoing different extents of conformational changes upon ligand binding. In all cases our protocol generates a significant fraction of structures featuring a low RMSD from the experimental holo geometry. Moreover, ensemble docking calculations using those conformations yielded in all cases native-like poses among the top-ranked ones.

Coupling enhanced sampling of the apo-receptor with template-based ligand conformers selection: performance in pose prediction in the D3R Grand Challenge 4

“…Each replica was simulated for 100 ns, leading to 400 ns of metadynamics simulations per window; coordinates were saved every 10 ps. Note that in [13] we demonstrated that EDES is not sensitive to the exact choice of the windows parameters. The height w of the Gaussian hills was set to 0.6 kcal/mol, while the widths si of the Gaussian hills were set to 0.06, 2.6, 1.7 and 3.0 respectively for RoGBS and CIP1,2,3, respectively.…”

“…In addition, the number of protein conformations was lowered to 200 to cope with the time constraints of the challenge, also considering that for each compound, 10 ligand conformations were employed in ensemble-docking runs. Protein conformations were extracted by means of a multi-step cluster analysis as described in [13], with the additional requirement to extract at least 10 cluster representatives from each of the 10 slices in which the RoGBS distributions were binned, so as to include a certain number of structures also from poorly sampled regions. The multi-step cluster analysis was applied separately to MDapo and EDES3w, extracting 500 clusters from each trajectory.…”

“…Here we report the performance of a new approach for ensemble-docking [11,12], which couples our recently proposed EDES (Ensemble-Docking with Enhanced-sampling of pocket Shape) protocol to sample holo-like and druggable conformations of proteins [13] by means of MD simulations [14,15] with the template-based algorithm for ligand conformer generation successfully employed in the previous GC3 competition [16]. EDES is peculiar in that, regarding the search for holo-like conformations of proteins, it exploits only the experimental structure of the apo-enzyme (PDB ID 1SGZ [17]).…”

“…Namely, EDES exploits an original set of collective variables (CVs) describing the geometry (that is, the shape and the volume) of the putative binding site(s). The method has been validated against targets undergoing very minor (single sidechain rearrangement) to very large (hinge-bending motions) conformational changes upon ligand binding [13]. In all cases, EDES was able to sample conformations of the binding site nearly identical to those occurring in the X-ray structures of the reference complexes, as well as to yield top ranked near-native docking poses, thus validating its potential as a new general approach to improve structure-based drug design.…”

Druggability Assessment in TRAPP using Machine Learning Approaches

“…However, proteins are dynamic and possess an inherent flexibility 17 which alters the shape and properties of their binding pockets. [18][19][20] Therefore, it is valuable to explore the different conformations of a binding pocket by molecular simulation and the corresponding variations in druggability. Some methods that combine pocket druggability prediction and molecular dynamics simulation have been developed, such as MDpocket 21 and JEDI.…”