Fig. 2: Docking specificity on Mtb DNA gyrase. | npj Drug Discovery

Fig. 2: Docking specificity on Mtb DNA gyrase.

From: AI-guided competitive docking for virtual screening and compound efficacy prediction

Fig. 2

A Molecular system: Left panel shows the full DNA gyrase system used for calculation; right panel provides a close-up of the binding site region. The three known binding sites are highlighted: FQ site (green), non-catalytic (NC) site (blue), and allosteric site (pink). The gyrase is displayed as a gray molecular surface and DNA is shown in orange. B AF3 clustering: AF3 successfully separates FQs (green), NC-site binders or NBTIs (blue), and allosteric inhibitors (pink) from unrelated compounds (orange and gray) based on pose convergence and proximity to the FQ site. Similar results were obtained with Boltz-1 (Table S3). C Boltz-2 clustering: In contrast, Boltz-2 fails to achieve this separation, clustering all binders near the FQ site. D-F Docking poses generated by AF3: Close-up views for FQs (D), NBTIs (E), and allosteric inhibitors (F). DF also presents a chart comparing AF3 pose convergence—defined as the percentage of true inhibitors and other ligands with pose convergence < 2.0 Å and a distance from the reference binding site < 5.0 Å—with Boltz-2 binding likelihood predictions, expressed as the percentage of inhibitors with a predicted binding likelihood > 0.5. A selected set of docked compounds is shown, including zoliflodacin (ZLF), trovafloxacin (TVF), AMK32b (ID8), gepotidacin (GPT), and thiophene 1 inhibitor (TH1). Each 3D image shows five representative AF3-predicted docking poses (i.e., the highest ipTM-scoring model for each seed) alongside the center of-mass of the reference FQ, MFX (gray sphere), superimposed on the reference crystal structure (PDB ID: 5BS8). Carbon atoms are color-coded as in (B), while oxygen and nitrogen atoms are shown in red and dark blue, respectively. The Mg2+ ions are represented as green spheres, the protein as a gray ribbon, and the DNA as an orange phosphate backbone with green–blue base sticks. Both the protein and DNA are shown in transparency. Protein visualizations were generated using PyMOL (The PyMOL Molecular Graphics System, Version 3.0, Schrödinger, LLC).

Back to article page