In the first stage, we will develop a structure-based CBLB-specific QSAR model for fast ligand-based screening based on the publicly available data of 3D CLBL-ligand structures and binding compounds. The distinguishing feature of the QSAR model is the graph neural network architecture coupled with Behler-Parrinello symmetry functions in the representation of protein-ligand complexes [1]. We will apply the derived QSAR model to billion-size REAL Enamine and obtain a focused chemical library to be tested with more computationally expensive approaches, as it follows.
In the second stage, we will compose a representative conformational ensemble of the CBLB TKB domain. We will consider available experimental structures as well as construct theoretical models using molecular mechanics approaches, such as molecular dynamics and non-linear normal mode analysis. We will analyze the obtained conformations using our deep learning binding profiling approach [2], resulting in the most promising binding site conformations for the structure-based virtual ligand screening.
In the third stage, given the focused chemical library and representative conformations obtained in the first and second stages, respectively, we will run 4D molecular docking using ICM-Pro docking suite (www.molsoft.com). We will use semi-empirical quantum mechanics calculations to generate starting 3D conformers of the focused chemical library and our custom protein preparation scripts to prepare the docking project [3]. We will use our deep learning-based scoring function [1] to estimate the binding affinity from the docked compounds. Thus, we will obtain the ranked list of the compounds, from which we will select the 100 hit candidates based on the score, chemical scaffold, and manual inspection of the binding poses.
[1] Karlov, Dmitry S., et al. "graphDelta: MPNN scoring function for the affinity prediction of protein–ligand complexes." ACS omega 5.10 (2020): 5150-5159.
[2] Kozlovskii, Igor, and Petr Popov. "Spatiotemporal identification of druggable binding sites using deep learning." Communications biology 3.1 (2020): 1-12.
[3] Grudinin, Sergei, et al. "Predicting binding poses and affinities in the CSAR 2013–2014 docking exercises using the knowledge-based Convex-PL potential." Journal of Chemical Information and Modeling 56.6 (2016): 1053-1062.