What is covalent docking? 

Introduction:

Covalent inhibitors are the drugs whose pharmacophores include an electrophilic “warhead” that forms a covalent bond with a target residues like cysteine, serine or lysine). Covalent inhibitors form an irreversible bond (though reversible covalent inhibitors exist) and locks the drug onto the target, resulting in very high potency and prolonged duration of action[1][2].

Historical Timeline of Key Covalent Drugs:

Serendipitous covalent drugs include:

• Aspirin: Which irreversibly acetylates the serine residue in the active site of the COX-1 enzyme, blocking prostaglandin synthesis.
• Penicillin: The beta-lactam ring acts as the covalent “warhead” that irreversibly binds to bacterial DD-transpeptidase, preventing cell wall synthesis. 
• Omeprazole: Forms a covalent disulfide bond with cysteines in the H+/K+ ATPase pump.
• Clopidogrel: Covalently and irreversibly binds to the P2Y12 receptor on platelets, preventing clotting.

Since the 2010s many rationally designed TCIs have reached the clinic: for instance, the BTK inhibitor ibrutinib (2013) and successors acalabrutinib/Zanubrutinib [3].  EGFR inhibitors afatinib/osimertinib for EGFR-mutant NSCLC (2013–15)[1], HCV protease inhibitors boceprevir/telaprevir (2011), KRAS G12C inhibitors sotorasib (2021) and adagrasib (phase III)[1], COVID-19 protease inhibitor nirmatrelvir (Paxlovid, 2021)[3].

Figure 1: Timeline showing the evolution of covalent drugs, from early medicines like aspirin and penicillin to modern targeted covalent inhibitors such as ibrutinib, dacomitinib, and zanubrutinib.

Success Stories:

Katritch. V et al describes the development of a covalent docking and virtual ligand screening (VLS) strategy to identify inhibitors of the vaccinia virus I7L cysteine protease, an enzyme essential for poxvirus replication. Because I7L is highly conserved among orthopoxviruses, including variola (smallpox) and monkeypox viruses, it represents an attractive antiviral drug target.

The major focus of the study was the implementation of a specialized covalent docking methodology within the MolSoft ICM platform. Since no experimental crystal structure of I7L was available, the researchers first generated a homology model using the yeast ubiquitin-like protease Ulp1 as a structural template. Although sequence identity between the proteins was low (~20%), the catalytic residues and active-site geometry were highly conserved, enabling accurate modeling of the binding pocket.

Unlike conventional docking approaches, the covalent docking protocol explicitly modeled the chemical reaction between ketone inhibitors and the catalytic cysteine residue (Cys328). Ketone warheads were transformed from planar SP2 geometry into tetrahedral SP3 thio-acyl intermediates to mimic the covalent transition state formed during nucleophilic attack by the cysteine thiol. Flexible tether restraints were applied to preserve the covalent geometry during docking simulations.

Using this approach, the authors performed covalent VLS of approximately 230,000 ketone and aldehyde compounds. Docking calculations evaluated van der Waals interactions, hydrogen bonding, electrostatics, desolvation, and entropy contributions. More than 900 compounds showed favorable docking scores, and 456 compounds were selected for biochemical testing.

Experimental validation using a fluorescence-based protease assay identified 97 active inhibitors, corresponding to an exceptionally high hit rate of approximately 21%. Several compounds exhibited micromolar potency, with the best inhibitor showing an IC50 near 6 µM. Aromatic ketones emerged as the most promising inhibitors because they balanced reactivity, specificity, and favourable drug-like properties[6].

This study demonstrated that covalent docking and covalent VLS can successfully identify cysteine protease inhibitors even without an experimentally determined protein structure, validating the effectiveness of the Molsoft ICM covalent docking platform for antiviral drug discovery.

MolSoft ICM Enabled Structure-Based Discovery of the FDA-Approved Drug Vabomere:

Scott J. Hecker et al. led to discovery and development of compound 9f (vaborbactam), a cyclic boronic acid β-lactamase inhibitor designed to restore the activity of carbapenem antibiotics against resistant Gram-negative bacteria. This work represents a major advancement in structure-based antibiotic discovery targeting carbapenem-resistant Enterobacteriaceae (CRE), particularly strains expressing Klebsiella pneumoniae carbapenemase (KPC).

The study focused on the rational design of a novel cyclic α-acylaminoboronic acid scaffold capable of inhibiting serine β-lactamases through reversible covalent interactions. Using structure-based modeling and docking within the MolSoft ICM platform, the researchers optimized inhibitor conformations to mimic the tetrahedral transition state formed during β-lactam hydrolysis. Covalent docking studies accurately predicted binding interactions with both class A and class C β-lactamases, demonstrating excellent agreement with subsequent X-ray crystal structures. The modelled covalent complex of compound 9f with AmpC β-lactamase showed an RMSD of only 0.8 Å relative to the experimental structure, validating the computational design strategy.

X-ray crystallography revealed that compound 9f forms substrate-like interactions within the enzyme active site. The boronate hydroxyl occupies the oxyanion hole, while additional hydrogen-bonding interactions stabilize binding across multiple β-lactamase classes. Importantly, the inhibitor displayed conformational adaptability, enabling potent inhibition of structurally distinct enzymes such as CTX-M-15 and AmpC.

Extensive microbiological profiling demonstrated that compound 9f strongly potentiated carbapenem antibiotics including meropenem and biapenem against resistant KPC-producing strains. In a neutropenic mouse lung infection model, combination therapy with 9f produced greater than a 2-log reduction in bacterial burden. Pharmacokinetic studies showed properties like β-lactam antibiotics, including high systemic exposure, short half-life, and low volume of distribution.

Safety studies demonstrated excellent tolerability, with no significant toxicity observed at high doses in preclinical studies or during Phase I clinical trials in healthy volunteers. The compound was later clinically developed as Vabomere (meropenem/vaborbactam), an FDA-approved therapy for complicated urinary tract infections caused by multidrug-resistant Gram-negative pathogens.

Figure 4: Predicted binding mode of RPX-7009 (Vabomere)

Benchmarking Studies:

The article represents a large-scale comparative benchmark of modern covalent docking programs used in structure-based drug discovery, with particular emphasis on the performance of MolSoft ICM-Pro. The study evaluated six major docking platforms—ICM-Pro, CovDock, AutoDock4, GOLD, FITTED, and MOE—using a curated dataset of 207 covalent protein–ligand crystal structures spanning multiple target classes, including kinases, proteases, and GTPases. The benchmark included diverse covalent warheads such as Michael acceptors, aldehydes, ketones, nitriles, and disulfide-forming ligands. 

Among all tested methods, ICM-Pro demonstrated the best overall docking accuracy and robustness. Using a 2.0 Å RMSD cutoff, ICM-Pro achieved approximately 62% success in predicting the correct Top1 binding pose and nearly 88–93% success within Top10 poses across several target classes. The platform showed especially strong performance for kinase inhibitors and Michael acceptor chemistries, making it highly suitable for covalent virtual ligand screening (VLS). 

Figure 5: Comparision of Covalent docking accuracy

A major advantage of ICM-Pro was its advanced covalent docking workflow. The software automatically generates pseudo-covalent ligand complexes, models stereochemical changes upon bond formation, and uses Monte Carlo conformational sampling combined with a full-atom scoring function optimized for covalent interactions. Unlike several competing methods, ICM-Pro maintained strong performance even with flexible ligands and large binding pockets. 

The study also demonstrated that accurate covalent docking depends heavily on non-covalent interaction networks, ligand flexibility, cysteine accessibility, and warhead geometry. ICM-Pro showed substantial improvement in prediction accuracy when multiple pharmacophoric interactions stabilized ligand binding, highlighting the importance of simultaneously modeling covalent and non-covalent interactions. 

This benchmarking established ICM-Pro as one of the most reliable and accurate platforms for covalent docking and virtual screening, validating its application in modern covalent drug discovery programs targeting reactive cysteine residues.

References:

1. Huang F. et al., “Covalent Warheads Targeting Cysteine Residue: The Promising Approach in Drug Development“ Molecules 27(22), 7728 (2022)
2. Namrashee V. et al., “The expanding repertoire of covalent warheads for drug discovery“ Drug Discovery Today 2023, Volume 28, Issue 12
3. Jesang Lee et al., “Extended Applications of Small-Molecule Covalent Inhibitors toward Novel Therapeutic Targets“ Pharmaceuticals 2022, 15(12), 1478
4. M.S. Hameed et al. “Advancements, challenges, and future frontiers in covalent inhibitors and covalent drugs: A review” European Journal of Medicinal Chemistry Reports 12 (2024) 100217
5. R Abagyan, M Totrov “Biased probability Monte Carlo conformational searches and electrostatic calculations for peptides and proteins” Journal of Molecular Biology. 1994 Jan 21;235(3):983-1002
6. Katritch V et al., “Discovery of small molecule inhibitors of ubiquitin-like poxvirus proteinase I7L using homology modeling and covalent docking approaches” J Comput Aided Mol Des (2007) 21:549–558
7. Scott J. Hecker et al., “Discovery of a Cyclic Boronic Acid β-Lactamase Inhibitor (RPX7009) with Utility vs Class A Serine Carbapenemases” J Med Chem. 2015 May 14;58(9):3682-92
8. Scarpino A. er al., “Comparative Evaluation of Covalent Docking Tools” J. Chem. Inf. Model. 2018, 58, 7, 1441–1458