Molecular Insights into Cyclooxygenase-2 Inhibitors Using Explainable Artificial Intelligence in Quantitative Structure–Activity Modeling

Hamid Bouseber; Mustapha Cherkaoui

doi:10.22146/ijc.117758

Molecular Insights into Cyclooxygenase-2 Inhibitors Using Explainable Artificial Intelligence in Quantitative Structure–Activity Modeling

https://doi.org/10.22146/ijc.117758

Hamid Bouseber^(1*), Mustapha Cherkaoui⁽²⁾

(1) Cadi Ayyad University, UCA, High Training Teachers School, LIRBEM Laboratory, BP 2400 Hay Hassani, 40000 Marrakesh, Morocco
(2) Cadi Ayyad University, UCA, High Training Teachers School, LIRBEM Laboratory, BP 2400 Hay Hassani, 40000 Marrakesh, Morocco
(*) Corresponding Author

Abstract

Cyclooxygenase-2 (COX-2) is an important therapeutic target for anti-inflammatory drug discovery. In this study, an integrated computational framework combining quantitative structure–activity relationship (QSAR) modeling, explainable artificial intelligence (XAI), molecular docking, and flexibility analysis was developed to investigate the structural determinants governing COX-2 inhibitory activity. A curated dataset of 2,297 compounds from the ChEMBL database was used to build QSAR regression models using PubChem, Morgan, and RDKit descriptors. Among the evaluated models, Random Forest combined with PubChem fingerprints achieved the best predictive performance and satisfactory robustness through cross-validation, external validation, and Y-Randomization analysis. Permutation importance and SHAP analyses identified aromatic, conjugated, and heteroatom-containing fragments as important contributors to inhibitory activity. The validated QSAR model was subsequently applied for virtual screening and prioritization of representative candidate compounds. Molecular docking and CABS-flex analyses further supported the stability of the selected ligand–COX-2 complexes. Overall, the proposed QSAR-XAI framework provides mechanistic insight into the molecular features associated with COX-2 inhibition and highlights the potential of interpretable machine learning for anti-inflammatory drug discovery.

Keywords

cyclooxygenase-2 (COX-2); QSAR; explainable artificial intelligence (XAI); SHAP; molecular docking

Full Text:

Full Text PDF

References

[1] Khalil, N.A., Ahmed, E.M., Tharwat, T., and Mahmoud, Z., 2024, NSAIDs between past and present; A long journey towards an ideal COX-2 inhibitor lead, RSC Adv., 14 (42), 30647–30661.

[2] Ahmadi, M., Bekeschus, S., Weltmann, K.D., von Woedtke, T., and Wende, K., 2022, Non-steroidal anti-inflammatory drugs: Recent advances in the use of synthetic COX-2 inhibitors, RSC Med. Chem., 13 (5), 471–496.

[3] Chahat, C., Gupta, S., Joshi, G., Wahajuddin, M., and Kumar, B., 2026, COX-2 inhibitors in inflammation and cancer: Recent developments in medicinal chemistry, Eur. J. Med. Chem., 310, 118781.

[4] Leal-Ramos, O.J., Arias-Ruiz, L.F., Huerta-Velázquez, J.M., Lamoreaux-Aguayo, J.P., Butcher, D., López-Cuellar, A.B., Orozco-Jiménez, K.I., and Torres-Bugarín, O., 2025, Mechanisms involved in the adverse cardiovascular effects of selective cyclooxygenase-2 inhibitors, Cardiovasc. Med., 28 (1), 5.

[5] Koirala, M., Yan, L., Mohamed, Z., and DiPaola, M., 2025, AI-integrated QSAR modeling for enhanced drug discovery: From classical approaches to deep learning and structural insight, Int. J. Mol. Sci., 26 (19), 9384.

[6] Cherkasov, A., Muratov, E.N., Fourches, D., Varnek, A., Baskin, I.I., Cronin, M., Dearden, J., Gramatica, P., Martin, Y.C., Todeschini, R., Consonni, V., Kuz’min, V.E., Cramer, R., Benigni, R., Yang, C., Rathman, J., Terfloth, L., Gasteiger, J., Richard, A., and Tropsha, A., 2014, QSAR modeling: Where have you been? Where are you going to?, J. Med. Chem., 57 (12), 4977–5010.

[7] Breiman, L., 2001, Random Forests, Mach. Learn., 45 (1), 5–32.

[8] Qadri, Y.A., Shaikh, S., Ahmad, K., Choi, I., Kim, S.W., and Vasilakos, A.V., 2025, Explainable artificial intelligence: A perspective on drug discovery, Pharmaceutics, 17 (9), 1119.

[9] Hossaini Alhashemi, S., Zare, F., Sadeghian, S., Poustforoosh, A., and Dehshahri, A., 2025, Identification of potential natural analgesic compounds through molecular docking-virtual screening, molecular dynamics simulation, MM/GBSA, DFT, and ADMET computations, Sci. Rep., 15 (1), 39995.

[10] Moussa, N., Hassan, A., and Gharaghani, S., 2021, Pharmacophore model, docking, QSAR, and molecular dynamics simulation studies of substituted cyclic imides and herbal medicines as COX-2 inhibitors, Heliyon, 7 (4), e06605.

[11] Rudrapal, M., Kirboga, K.K., Abdalla, M., and Maji, S., 2024, Explainable artificial intelligence-assisted virtual screening and bioinformatics approaches for effective bioactivity prediction of phenolic cyclooxygenase-2 (COX-2) inhibitors using PubChem molecular fingerprints, Mol. Diversity, 28 (4), 2099–2118.

[12] Ghosh, P., Kumar, A., and Singh, S.K., 2025, COX‐2 inhibitor prediction with KNIME: A codeless automated machine learning‐based virtual screening workflow, J. Comput. Chem., 46 (2), e70030.

[13] Zdrazil, B., Felix, E., Hunter, F., Manners, E.J., Blackshaw, J., Corbett, S., de Veij, M., Ioannidis, H., Lopez, D.M., Mosquera, J.F., Magarinos, M.P., Bosc, N., Arcila, R., Kizilören, T., Gaulton, A., Bento, A.P., Adasme, M.F., Monecke, P., Landrum, G.A., and Leach, A.R., 2024, The ChEMBL Database in 2023: A drug discovery platform spanning multiple bioactivity data types and time periods, Nucleic Acids Res., 52 (D1), D1180–D1192.

[14] Bento, A.P., Hersey, A., Félix, E., Landrum, G., Gaulton, A., Atkinson, F., Bellis, L.J., De Veij, M., and Leach, A.R., 2020, An open source chemical structure curation pipeline using RDKit, J. Cheminf., 12 (1), 51.

[15] Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, É., 2011, Scikit-learn: Machine learning in python, J. Mach. Learn. Res., 12 (85), 2825–2830.

[16] Eberhardt, J., Santos-Martins, D., Tillack, A.F., and Forli, S., 2021, AutoDock Vina 1.2.0: New docking methods, expanded force field, and python bindings, J. Chem. Inf. Model., 61 (8), 3891–3898.

[17] Kurcinski, M., Oleniecki, T., Ciemny, M.P., Kuriata, A., Kolinski, A., and Kmiecik, S., 2019, CABS-flex standalone: A simulation environment for fast modeling of protein flexibility, Bioinformatics, 35 (4), 694–695.

[18] Tropsha, A., 2010, Best practices for QSAR model development, validation, and exploitation, Mol. Inf., 29 (6-7), 476–488.

[19] O’Boyle, N.M., Banck, M., James, C.A., Morley, C., Vandermeersch, T., and Hutchison, G.R., 2011, Open Babel: An open chemical toolbox, J. Cheminf., 3 (1), 33.

[20] Li, Z., Wang, Z.C., Li, X., Abbas, M., Wu, S.Y., Ren, S.Z., Liu, Q.X., Liu, Y., Chen, P.W., Duan, Y.T., Lv, P.C., and Zhu, H.L., 2019, Design, synthesis and evaluation of novel diaryl-1,5-diazoles derivatives bearing morpholine as potent dual COX-2/5-LOX inhibitors and antitumor agents, Eur. J. Med. Chem., 169, 168–184.

[21] Ren, S.Z., Wang, Z.C., Zhu, X.H., Zhu, D., Li, Z., Shen, F.Q., Duan, Y.T., Cao, H., Zhao, J., and Zhu, H.L., 2018, Design and biological evaluation of novel hybrids of 1, 5-diarylpyrazole and Chrysin for selective COX-2 inhibition, Bioorg. Med. Chem., 26 (14), 4264–4275.

[22] Kim, S., Chen, J., Cheng, T., Gindulyte, A., He, J., He, S., Li, Q., Shoemaker, B.A., Thiessen, P.A., Yu, B., Zaslavsky, L., Zhang, J., and Bolton, E.E., 2021, PubChem in 2021: New data content and improved web interfaces, Nucleic Acids Res., 49 (D1), D1388–D1395.

[23] Yap, C.W., 2011, PaDEL-descriptor: An open source software to calculate molecular descriptors and fingerprints, J. Comput. Chem., 32 (7), 1466–1474.

[24] Nguyen-Vo, T. H., Teesdale-Spittle, P., Harvey, J.E., and Nguyen, B.P., 2024, Molecular representations in bio-cheminformatics, Memetic Comput., 16 (3), 519–536.

[25] Moslemi, A., 2023, A tutorial-based survey on feature selection: Recent advancements on feature selection, Eng. Appl. Artif. Intell., 126, 107136.

[26] Friedman, J.H., 2001, Greedy function approximation: A gradient boosting machine, Ann. Stat., 29, 1189–1232.

[27] De, P., Kar, S., Ambure, P., and Roy, K., 2022, Prediction reliability of QSAR models: An overview of various validation tools, Arch. Toxicol., 96 (5), 1279–1295.

[28] Li, J., Zhao, T., Yang, Q., Du, S., and Xu, L., 2025, A review of quantitative structure-activity relationship: The development and current status of data sets, molecular descriptors and mathematical models, Chemom. Intell. Lab. Syst., 256, 105278.

[29] Mora, J.R., Marquez, E.A., Pérez-Pérez, N., Contreras-Torres, E., Perez-Castillo, Y., Agüero-Chapin, G., Martinez-Rios, F., Marrero-Ponce, Y., and Barigye, S.J., 2024, Rethinking the applicability domain analysis in QSAR models, J. Comput.-Aided Mol. Des., 38 (1), 9.

[30] Han, H., Shaker, B., Lee, J.H., Choi, S., Yoon, S., Singh, M., Basith, S., Cui, M., Ahn, S., An, J., Kang, S., Yeom, M.S., and Choi, S., 2025, Employing automated machine learning (AutoML) methods to facilitate the in silico ADMET properties prediction, J. Chem. Inf. Model., 65 (7), 3215–3225.

[31] Lundberg, S.M., and Lee, S.I., 2017, “A unified approach to interpreting model predictions” in Advances in Neural Information Processing Systems, Curran Associates, Inc., Red Hook, NY, US, 4768–4777.

[32] Paggi, J.M., Pandit, A., and Dror, R.O., 2024, The art and science of molecular docking, Annu. Rev. Biochem., 93, 389–410.

[33] Ma, Z., Ajibade, A., and Zou, X., 2024, Docking strategies for predicting protein-ligand interactions and their application to structure-based drug design, Commun. Inf. Syst., 24 (3), 199–230.

[34] Dablander, M., Hanser, T., Lambiotte, R., and Morris, G.M., 2024, Sort and slice: A simple and superior alternative to hash-based folding for extended-connectivity fingerprints, J. Cheminf., 16 (1), 135.

DOI: https://doi.org/10.22146/ijc.117758