Now Available on Superbio.ai

April 14, 2024

<aside> 🔖 SieveStack’s proprietary tool for covalent modeling is now available through our partner,  Superbio.ai.

</aside>

Covalent inhibitors have made a remarkable comeback in recent drug discovery — targeting KRAS(G12C), EGFR, BTK, and SARS-CoV-2’s main protease, to name a few. These molecules typically feature reactive functional groups (warheads) that form covalent bonds with target residues, improving potency and residence time over reversible compounds. [1]

Challenges in Covalent Modeling

• Reactive warheads must be positioned close to a suitable nucleophile
• Non-covalent portions of the ligand must be stable and well-optimized
• Reaction geometry must allow for feasible covalent bond formation
• Protein flexibility is critical to capture realistic binding modes

At SieveStack, we’ve developed an internal toolchain to address these requirements. It integrates physics-based simulation, flexible side-chain modeling, and automated scoring to evaluate covalent binding poses. To support broader use, the tool has now been deployed on Superbio.ai’s cloud platform.

Examples

To demonstrate the utility of our tool, we applied it to several well-characterized covalent inhibitors. In these cases, the predicted ligand poses closely matched the experimental co-crystal structures, with RMSD values under 1Å. The following examples highlight accurate reproduction of covalent binding modes across high-value targets, consistent with crystallographic data.

Legend

Target: BTK Covalent Ligand: Zanubrutinib PDB Code: 6J6M

Target: SARS-CoV-2 Main Protease Covalent Ligand: Nirmatrelvir PDB Code: 8DZ2