An Integrative Machine Learning Framework for Drug Repurposing Using Molecular Docking and Drug Sentiment Analysis in Type 2 Diabetes

or control drug for the molecular docking analysis. The prepared structures were then saved in PDB format for subsequent docking analysis. Figure 4 Bupropion P-ISSN 2695-222X Figure 5 Adapalene P-ISSN 2695-222X Figure 6 Esomeprazole P-ISSN 2695-222X Figure 7 Sitagliptin 2.3.2. Protein Selection and Preparation Dipeptidyl peptidase-4 (DPP-4), a well-known target in type 2 diabetes was selected as the protein target for the docking analysis. The activity of DPP-4 can be inhibited by some medications (DPP- 4 inhibitors) used to manage high blood glucose levels in patients with type 2 diabetes mellitus. Some of these inhibitors are; Sitagliptin, Saxagliptin, Vildagliptin, Alogliptin, Linagliptin, Denagliptin, Dutogliptin and Omarigliptin (Mohd, 2024). The three-dimensional (3d) structure of the protein target (DPP-4) was obtained from the Protein Data Bank . The structure was cleaned and prepared in PyMOL in the following six (6) steps: 1. Confirmation of 3d structure 2. Removal of Water molecule 3. Retaining Carbohydrate and Polymer structure 4. Removal of N7F ligand 5. Addition of hydrogen 6. Download and save as pdb format, which is the acceptable format for docking on CB-Dock2 Protein–ligand docking was performed using CB-Dock2. CB-Dock2 is an automated blind- docking platform that combines binding-cavity detection with docking and template-based information to identify potential binding sites and ligand poses (Liu et al., 2022). Its docking workflow incorporates AutoDock Vina, a widely used molecular-docking program that estimates ligand binding poses and provides a scoring function for ranking predicted complexes (Sarkar et al., 2024). For each protein–ligand complex, the best-ranked pose, predicted Vina docking score, and docking-box dimensions were recorded for comparison. P-ISSN 2695-222X Figure 8 DPP-4 3. Results and Discussion 3.1 Sentiment Analysis The Table 2 below shows the sentiment analysis score of the three selected drug candidates and molecular docking results for the three drug candidates and the control drug, Sitagliptin. The sentiment score of 1.0 was achieved for Welbutrin-XL (Bupropion), and Nexium (Esomeprazole), while a sentiment score of 0.84 was achieved for Differin (Adapalene). A positive sentiment score can be interpreted as a favourable patient reported experience with the drugs but not evidence of therapeutic efficacy for type 2 diabetes. Therefore, sentiment analysis of drug reviews alone cannot establish the antidiabetic potential of the drugs. 3.2 Molecular Docking Analysis The Protein-ligand blind docking was performed on CB-Dock2, a protein-ligand blind docking server that integrates both structure-based cavity detection with template based molecular docking methods (Yang et at., 2022). The docking analysis shows that all the four ligands (including the control drug) had the best ranked pose in cavity C3. DPP4-Sitagliptin complex (control) had a vina score of -9.3 kcal/mol and dimension 25 x 25 x 25, DPP4-Bupropion complex had a vina score of −6.3 kcal/mol and dimension 19 x 19 x 19, DPP4-Esomeprazole complex had a via score of −8.8 kcal/mol and dimension 23 x 23 x 23, and DPP4-Adapalene complex produced a Vina score 0f −10.8 kcal/mol and dimension 24 x 24 x 24. Sitagliptin recorded the largest dimension. The dimensions describe the docking search space and should not be interpreted as measures of binding strength. Sitagliptin is one of the prominent DPP-4 inhibitors, exhibiting potent hypoglycemic effects. Although it is associated with some side effects including; gastrointestinal issues, rashes, and more (Huang et al., 2020). P-ISSN 2695-222X Compared with the reference drug, Sitagliptin, Adapalene produced a more negative predicted Vina docking score (−10.8 kcal/mol versus −9.3 kcal/mol). This suggests that Adapalene may have a favorable predicted interaction with DPP-4. However, additional computational and experimental studies would be required to determine whether this predicted interaction translates into DPP-4 inhibitory activity. In this study, Sitagliptin remains the established DPP-4 inhibitor used as the reference compound. Drug/Ligand Protein Target Sentiment Score Vina Docking Score (kcal/mol) Best Pose Docking Dimensions (X × Y × Z) Sitagliptin (Control) DPP-4 - -9.3 C3 25 x 25 x 25 Wellbutrin-XL (Bupropion) DPP-4 1.0 -6.3 C3 19 × 19 × 26 Nexium (Esomeprazole) DPP-4 1.0 -8.8 C3 23 × 23 × 23 Differin (Adapalene) DPP-4 0.84 -10.8 C3 24 × 24 × 24 Table 2 Sentiment Analysis and Molecular docking result P-ISSN 2695-222X Figure 9 DPP-4-Adapalene Complex Figure 10 DPP-4-Sitagliptin Complex 4. Limitations of the Research There are several limitations that should be considered when interpreting this research. First, the sentiment analysis was based on drug reviews obtained from the UCI Machine Learning Repository under the Drug Reviews (Druglib.com) dataset, and the experiences reported by patients may relate to different medical conditions and therapeutic indications. Therefore, the P-ISSN 2695-222X sentiment scores do not necessarily represent patient experiences with the use of these drugs for type 2 diabetes. Second, molecular docking provides computational predictions of ligand–protein interactions and does not by itself establish target inhibition, pharmacological activity, safety, or clinical efficacy. Finally, the study did not include molecular dynamics simulations, biochemical assays, cell-based experiments, or clinical studies to experimentally validate the predicted interactions. Therefore, the identified compounds should be regarded as computationally prioritized candidates which require further investigation rather than confirmed therapeutic agents for type 2 diabetes. 5. Conclusion This study developed an integrative computational framework that combined drug-review sentiment analysis with molecular docking to prioritize non-diabetes drugs as potential therapeutic candidates for type 2 diabetes. Logistic Regression was the best-performing sentiment- classification model based on macro recall (0.567231) and macro F1 score (0.558643). Three non- diabetes drugs; Bupropion, Esomeprazole, and Adapalene were subsequently selected based on their positive sentiment scores and evaluated against DPP-4 using CB-Dock2. Sitagliptin was used as the control drug and also docked with DPP-4 on CB-Dock2. Among the evaluated compounds, Adapalene produced the most favorable predicted Vina docking score (−10.8 kcal/mol), compared with -9.3 kcal/mol for Sitagliptin, −8.8 kcal/mol for Esomeprazole and −6.3 kcal/mol for Bupropion. These findings suggest that adapalene warrants further investigation as a computationally prioritized candidate for DPP-4-related studies. However, the results do not establish antidiabetic efficacy or clinical suitability. Further validation through repeated or consensus docking, molecular dynamics simulations, biochemical assays, and appropriate experimental studies is required before any therapeutic conclusion can be drawn. Overall, the framework demonstrates how patient-generated textual evidence and structure-based computational evidence can be combined to support the prioritization of existing drugs for future drug-repurposing research. P-ISSN 2695-222X References American Diabetes Association Professional Practice Committee for Diabetes. (2026). 1. Improving care and promoting health in populations: Standards of Care in Diabetes—2026. Diabetes Care, 49(Supplement 1), S13–S26. https://doi.org/10.2337/dc26-S001 Cousins, H. C., Nayar, G., & Altman, R. B. (2024). Computational approaches to drug repurposing: Methods, challenges, and opportunities. Annual Review of Biomedical Data Science, 7, 15–29. https://doi.org/10.1146/annurev-biodatasci-110123-025333 Dwivedi, S. K., & Rawat, B. (2015). A review paper on data preprocessing: A critical phase in web usage mining process. In Proceedings of the 2015 International Conference on Green Computing and Internet of Things (ICGCIoT) (pp. 506–510). IEEE. https://doi.org/10.1109/ICGCIoT.2015.7380517 Genitsaridi, I., Salpea, P., Salim, A., Sajjadi, S. F., Tomic, D., James, S., Thirunavukkarasu, S., Issaka, A., Chen, L., Basit, A., Luk, A. O. Y., Ma, R. C. W., Mbanya, J.-C., Ramachandran, A., Wild, S. H., Duncan, B. B., Boyko, E. J., & Magliano, D. J. (2026). 11th edition of the IDF Diabetes Atlas: Global, regional, and national diabetes prevalence estimates for 2024 and projections for 2050. The Lancet Diabetes & Endocrinology, 14(2), 149–156. https://doi.org/10.1016/S2213-8587(25)00299-2 Huang, J., Jia, Y., Sun, S., & Meng, L. (2020). Adverse event profiles of dipeptidyl peptidase-4 inhibitors: Data mining of the public version of the FDA adverse event reporting system. BMC Pharmacology and Toxicology, 21(1), 68. https://doi.org/10.1186/s40360-020- 00447-w Liu, Y., Yang, X., Gan, J., Chen, S., Xiao, Z.-X., & Cao, Y. (2022). CB-Dock2: Improved protein– ligand blind docking by integrating cavity detection, docking and homologous template fitting. Nucleic Acids Research, 50(W1), W159–W164. https://doi.org/10.1093/nar/gkac394 Mendez, D., Gaulton, A., Bento, A. P., Chambers, J., De Veij, M., Félix, E., Magariños, M. P., Mosquera, J. F., Mutowo, P., Nowotka, M., Gordillo-Marañón, M., Hunter, F., Junco, L., Mugumbate, G., Rodriguez-Lopez, M., Atkinson, F., Bosc, N., Radoux, C. J., Segura- Cabrera, A., Hersey, A., & Leach, A. R. (2019). ChEMBL: Towards direct deposition of bioassay data. Nucleic Acids Research, 47(D1), D930–D940. https://doi.org/10.1093/nar/gky1075 Naim, M. J. (2024). A review of dipeptidyl peptidase-4 (DPP-4) and its potential synthetic derivatives in the management of diabetes mellitus. Journal of Angiotherapy, 8(1), 1–12. https://doi.org/10.25163/angiotherapy.819417 Nivatya, H. K., Singh, A., Kumar, N., Sonam, Sharma, L., Singh, V., Mishra, R., Gaur, N., & Mishra, A. K. (2025). Assessing molecular docking tools: Understanding drug discovery and design. Future Journal of Pharmaceutical Sciences, 11, 111. https://doi.org/10.1186/s43094-025-00862-y Sarkar, A., Concilio, S., Sessa, L., Marrafino, F., & Piotto, S. (2024). Advancements and novel approaches in modified AutoDock Vina algorithms for enhanced molecular docking. Results in Chemistry, 7, 101319. https://doi.org/10.1016/j.rechem.2024.101319 Vamathevan, J., Clark, D., Czodrowski, P., Dunham, I., Ferran, E., Lee, G., Li, B., Madabhushi, A., Shah, P., Spitzer, M., & Zhao, S. (2019). Applications of machine learning in drug discovery and development. Nature Reviews Drug Discovery, 18(6), 463–477. https://doi.org/10.1038/s41573-019-0024-5 P-ISSN 2695-222X Yang, X., Liu, Y., Gan, J., Xiao, Z.-X., & Cao, Y. (2022). FitDock: Protein–ligand docking by template fitting. Briefings in Bioinformatics, 23(3), bbac087. https://doi.org/10.1093/bib/bbac087