3D pharmacophore-based virtual screening, molecular dynamics simulations, and high-throughput docking

In our workflow we will employ 3D-pharmacophore screening to synergize the information coming from co-crystalized fragment crystal structures with the information from molecular dynamics simulations.

Firstly, we will capture the possible protein-ligand interactions in the ADPr binding site of SARS-CoV-2 Mac1 through apo-site molecular dynamics (MD) simulations and PyRod [1]. PyRod is an open-source software that tracks water molecules in MD simulations. Through that process, it identifies hydrogen bonds, as well as hydrophobic, charged, and aromatic areas in the binding site. This information is contained in dynamic molecular interaction fields (dMIFs), that can be visualized or applied to generate pharmacophore features. We will visualize the dMIFs as grids in LigandScout to evaluate their binding contribution and steer our selection of the pharmacophore features in the binding site (PyRod’s “super pharmacophore”) [2].

Secondly, we aim to imitate protein-ligand interactions associated to high ligand efficacy, as shown from a massive crystallographic fragment screen. To apply these pharmacophore features in their induced-fit arrangement, we will first align the crystal structures of fragments occupying the adenine subsite with MOE and use LigandScout to derive structure-based pharmacophores from them [3]. Integrated 3D pharmacophores will then be generated by merging the pharmacophore features originating from both crystal structures (LigandScout) and molecular dynamics simulations (PyRod).

To limit the extensive chemical space of the Enamine REAL collection and rationalize the use of our computational resources, a substructure search will be performed with RDKit to prepare a library of compounds capable of forming interactions derived from the structure-based pharmacophores. These compounds will be ionized using MOE.

The integrated pharmacophore models will then be employed for a virtual screen and subsequent pharmacophore filtering of the docked poses generated with GOLD. Docked poses of the molecules will be energy-minimized in the binding site and visually inspected to ensure shape complementarity to the binding site and their diversity (eliminating compounds through pairs with the highest Tanimoto index).

In the hit optimization phase we will follow the same workflow, but instead of considering the co-crystalized fragments, we will consider the docked pose of our experimentally validated hit compounds from the hit identification phase. Since PyRod is implicated in discovering and evaluating novel interactions in the apo binding site, we will replace it with the use of our open-source software OpenMMDL. This will allow us to evaluate the stability of the docked poses and to analyze the possible binding mechanisms dynamically.

The combination of our two structure-based approaches, being the application of the in-house program Pyrod for the creation of 3D pharmacophores based upon molecular dynamics simulations of water molecules in the protein binding cavities in addition to the use of 3D pharmacophores of the co-crystallized ligands would allow us to generate 3D pharmacophores with certain features, that would increase the chance of obtaining high-affinity ligands.

Dynamic 3D Pharmacophores

InteLigand - LigandScout

CCG - MOE

Schrodinger - Desmond 

CCDC - GOLD

 

PyRod

OpenMMDL

RDKit

KNIME

Python

[1] Schaller D, Pach S, Wolber G. PyRod: Tracing Water Molecules in Molecular Dynamics Simulations. J Chem Inf Model. 2019 Jun 24;59(6):2818-29. (10.1021/acs.jcim.9b00281)

[2] Pach S, Sarter TM, Yousef R, Schaller D, Bergemann S, Arkona C, et al. Catching a Moving Target: Comparative Modeling of Flaviviral NS2B-NS3 Reveals Small Molecule Zika Protease Inhibitors. ACS Med Chem Lett. 2020 Apr 9;11(4):514-20. (10.1021/acsmedchemlett.9b00629)

[3] Schaller D, Šribar D, Noonan T, Deng L, Nguyen TN, Pach S, et al. Next generation 3D pharmacophore modeling. WIREs Comput Mol Sci. 2020 Jul;10(4) (10.1002/wcms.1468)

3D pharmacophore models of each of the most active hits will be generated. These pharmacophores might show additional interaction points not included in the starting pharmacophore model. The binding affinity data will be related to the static 3D pharmacophores of the hit compounds to generate initial structure-activity relationship models [1]. These will be further investigated using our in-house dynamic pharmacophore (dynophore) method [2-6], based on molecular dynamics simulations of the ligands within the NSP3 binding cavity.  The fully-automated dynophore method analyzes the interaction pattern between ligand and protein over the course of MD simulations, generating a visual representation of the types of 3D pharmacophore feature types that occur over the course of the simulation, along with their respective statistical occurrences.

Dynophores from most active compounds will be geometrically merged and chemical features weights will be trained using active compounds using different machine learning techniques. Furthermore, water-mediated interactions between ligand and protein will be  analyzed to provide further insight into the interactions contributing to binding affinity. These analyses will determine the contribution of different interactions to ligand binding affinities, suggesting which chemical and structural properties to prioritize in the derivative compounds. Derivatives will be identified via two different methods. One will be to perform a shape-based analog search using the ROCS software from OpenEye [7]. The other will be the combinatorial in-silico generation of active ligand-derivatives using the Reactor software from ChemAxon (https://chemaxon.com/products/reactor) with subsequent docking and dynophore-based scoring.

The Application of our in-house dynamic pharmacophore (dynophore) method [2-6] will allow us to analyze the interaction patterns between the ligand and the corresponding protein over the course of the molecular dynamics simulations and the obtained data from the analysis would be compared with the experimental data. Additionally different machine learning methods would be applied to train the geometrically merged dynophores and chemical features of the active compounds, thus allowing us to purpose chemical and structural properties in the optimization of the compounds.

Dynamic interaction pattern-based SAR determination followed by shape-based chemical similarity search and in-silico combinatorial derivative generation.

InteLigand - LigandScout
CCG - MOE
Schrodinger - Desmond
OpenEye - ROCS
ChemAxon - Reactor

Dynophore

1. Pach S, Sarter TM, Yousef R, Schaller D, Bergemann S, Arkona C, et al. Catching a Moving Target: Comparative Modeling of Flaviviral NS2B-NS3 Reveals Small Molecule Zika Protease Inhibitors. ACS Med Chem Lett. 2020 Apr 9;11(4):514-20. (10.1021/acsmedchemlett.9b00629)

2. Bock A, Bermudez M, Krebs F, Matera C, Chirinda B, Sydow D, et al. Ligand Binding Ensembles Determine Graded Agonist Efficacies at a G Protein-coupled Receptor. Journal of Biological Chemistry. 2016 Jul;291(31):16375-89.(10.1074/jbc.M116.735431)

3. Machalz D, Pach S, Bermudez M, Bureik M, Wolber G. Structural insights into understudied human cytochrome P450 enzymes. Drug Discovery Today. 2021 Oct;26(10):2456-64. (10.1016/j.drudis.2021.06.006) 

4. Machalz D, Li H, Du W, Sharma S, Liu S, Bureik M, et al. Discovery of a novel potent cytochrome P450 CYP4Z1 inhibitor. European Journal of Medicinal Chemistry. 2021 Apr;215:113255. (10.1016/j.ejmech.2021.113255)

5. Denzinger K, Nguyen TN, Noonan T, Wolber G, Bermudez M. Biased Ligands Differentially Shape the Conformation of the Extracellular Loop Region in 5-HT2B Receptors. IJMS. 2020 Dec 20;21(24):9728. (10.3390/ijms21249728)

6. Meşeli T, Doğan ŞD, Gündüz MG, Kökbudak Z, Skaro Bogojevic S, Noonan T, et al. Design, synthesis, antibacterial activity evaluation and molecular modeling studies of new sulfonamides containing a sulfathiazole moiety. New J Chem. 2021;45(18):8166-77. (10.1039/d1nj00150g) 

7. Šribar D, Grabowski M, Murgueitio MS, Bermudez M, Weindl G, Wolber G. Identification and characterization of a novel chemotype for human TLR8 inhibitors. European Journal of Medicinal Chemistry. 2019 Oct;179:744-52. (10.1016/j.ejmech.2019.06.084)