Development and Application of CDOCKER Docking Methodology

Abstract

The binding of small molecule ligands to receptor targets is important to numerous biological processes. When multiple ligands with different sizes are docked to a target receptor, it is reasonable to assume the residues in the binding pocket may adopt alternative conformations. It has also been suggested that the entropic contribution to binding can be important. In the work presented here, we discuss a new physics-based scoring function that includes both enthalpic and entropic contributions to binding that consider the conformational variability of the flexible receptor side chains within the ensemble of docked poses. To accommodate the additional searching requirements for flexible receptor docking studies, we also describe a novel hybrid searching algorithm that combines both molecular dynamics (MD) based simulated annealing and a continuous genetic algorithm. These advances further the development of the Flexible CDOCKER docking methods in the CHARMM software package. We benchmark our developments using 6 varied receptor targets: thrombin, dihydrofolate reductase (DHFR), T4 L99A, T4 L99A/M102Q, PDE10A and the riboswitch. These cover a wide range of enzyme classes with different binding pocket environments and the novel aspects of RNA receptor targets. We demonstrate improved accuracy in flexible cross-docking experiments compared with rigid cross-docking. The largest improvement in top ranking accuracy is 22% for ligands binding to DHFR. Moreover, by using advances in GPU accelerated sampling, we reduce the wall-time required for such calculations so that flexible receptor ligand docking can be applied in high-throughput experiments.

As a practical example of this methodology, we worked with a team of experimental colleagues to identify potential therapeutics for the host transmembrane serine protease TMPRSS2, a promising antiviral target that plays a direct role in SARS-CoV-2 infections. We screened a database of 134,109 molecules that were collected from multiple sources of molecules in clinical trials, on approved drug lists or readily available in local libraries. These compounds were filtered based on pharmacophore similarity. A total of 4,308 candidates were identified and docked with the flexible docking method noted above. From our in silico docking trials, novel non-covalent inhibitors were identified and verified with biochemical assays.

Targeted covalent inhibitors (TCIs) are an important component in the toolbox of drug discovery and about 30% of currently marketed drugs are TCIs. Although these drugs raise concerns about toxicity, their high potencies and prolonged effects result in less-frequent drug dosing and wide therapeutic margins for patients. On the other hand, a recent study shows that one of the known inhibitors, Nafamostat, forms a covalent bond with TMPRSS2. This motivated us to expand the application of Rigid CDOCKER to tethered docking. We describe the implementation and testing of a covalent docking methodology in Rigid CDOCKER and the optimization of the corresponding physics-based scoring function with an additional customizable covalent bond grid potential which represents the free energy change of bond formation between the ligand and the receptor. We show that our new covalent docking algorithm has comparable pose prediction accuracy with reduced computational cost and could identify lead compounds among a large chemical space.

Finally, leveraging the popularity and utility of the Python language, we introduce the python functionality in the CDOCKER family (i.e., pyCHARMM CDOCKER). This further accelerates CDOCKER docking algorithms and broadens the potential users who want to perform standard CDOCKER docking experiments with little knowledge of docking or the CDOCKER family.