Benzodiazepines docked into human GABAAR and schistosome TRP channel binding pockets

Modeling benzodiazepine binding to TRP and GABAARs

Targets were obtained from public structure repositories. GABAAR coordinates were taken from RSCB, PDB accession 6HUP for a cryo-EM derived structure in its diazepam-bound state. The AlphaFold model AF-A0A5K4FCC0-F1-model provided starting coordinates for the S. mansoni TRP channel Smp_333650. Structure models were prepared for docking using the DockPrep utility of ChimeraX v1.7 (Meng et al. 2023) to fill in missing atoms, add hydrogens, assign partial charges (AMBER ff14SB), and exported in mol2 formats. SMILES for the benzodiazepine chemical series were generated using ChemDraw. Library preparation utilities fixpka (QUACPAC v2.1.2.1), OMEGA2 v4.1.1.1, and molcharge (QUACPAC v2.1.2.1) were used to assign protonation states, build energetically reasonable starting 3D conformers for docking, and assign partial charges (MMFF), respectively (OpenEye Scientific, Cadence Molecular Sciences, Inc.). Single conformations were exported in mol2 format for docking. Docking was performed using the program Gnina v1.0 (McNutt et al. 2021). To localize the search space in docking GABAAR, the binding pocket was specified based on the position of diazepam using the autobox option with a bounding box of 10 Angstroms. No sidechain flexibility was allowed for docking into the GABAAR. For the schistosome TRP channel, the site center was determined using the most probable site in the VSLD domain using the prediction tool p2rank (Krivák and Hoksza 2018). Due to steric restrictions within the putative binding pocket of the TRP channel, sidechain flexibility was enabled for pocket residues H827, H831, Y864, E867, E868, D893, F940, R943, H946, I947, Y1117, R1120. For both targets, extensive sampling was achieved using an exhaustiveness parameter of 64. CNN scoring was used for re-scoring final poses by setting cnn_scoring option to “rescore.” For CNN-based pose scoring, the “crossdock_default2018” model was selected. Docking poses were inspected and images created using PyMOL v2.5.4 (Schrodinger, LLC).

Krivák, Radoslav, and David Hoksza. 2018. “P2Rank: Machine Learning Based Tool for Rapid and Accurate Prediction of Ligand Binding Sites from Protein Structure.” Journal of Cheminformatics 10 (1): 39.

McNutt, Andrew T., Paul Francoeur, Rishal Aggarwal, Tomohide Masuda, Rocco Meli, Matthew Ragoza, Jocelyn Sunseri, and David Ryan Koes. 2021. “GNINA 1.0: Molecular Docking with Deep Learning.” Journal of Cheminformatics 13 (1): 43.

Meng, Elaine C., Thomas D. Goddard, Eric F. Pettersen, Greg S. Couch, Zach J. Pearson, John H. Morris, and Thomas E. Ferrin. 2023. “UCSF ChimeraX: Tools for Structure Building and Analysis.” Protein Science: A Publication of the Protein Society 32 (11): e4792.