Adverse drug events (ADEs) remain a leading cause of morbidity, hospitalization, and preventable mortality across healthcare systems worldwide. Traditional pharmacovigilance systems rely on spontaneous reporting and post-marketing surveillance, which are limited by underreporting, latency, and fragmented data integration. The growing availability of heterogeneous biomedical data—including chemical structures, biological targets, electronic health records (EHRs), spontaneous reporting systems, and pharmacogenomic profiles—has catalyzed the development of machine learning (ML) approaches for predictive pharmacovigilance.

This study provides a comprehensive integrative research framework for predictive modeling of drug side effects and ADEs by synthesizing similarity-based, network-based, ensemble, deep learning, and explainable artificial intelligence (XAI) approaches. It further evaluates translational considerations, external validation strategies, pediatric and vulnerable populations, and regulatory alignment for software as a medical device.

Drawing exclusively on established literature, we construct a unified methodological architecture integrating drug similarity models, multi-label ensemble learning, graph neural networks, EHR-based prediction, signal detection in spontaneous reporting systems, and model validation frameworks. Theoretical synthesis is conducted across molecular databases (e.g., SIDER, PubChem), multi-site EHR studies, and contemporary machine learning paradigms. Reporting and validation frameworks are aligned with TRIPOD and TRIPOD-SRMA guidelines, and regulatory guidance for AI-based medical software is incorporated.

The integrative model demonstrates conceptual advantages in capturing multi-dimensional drug–target–phenotype interactions. Evidence across prior studies supports improved discrimination using ensemble and hybrid deep learning approaches, particularly in hemorrhage, nephrotoxicity, cardiotoxicity, and immune-related adverse events. Explainable methods enhance transparency by identifying clinically interpretable predictors, while external validation remains a critical determinant of generalizability.

Adverse drug events, Machine learning, Pharmacovigilance, Electronic health records, Explainable AI, Drug safety prediction

Al-Dhaher Z, et al. Activating and tranquilizing effects of first-time treatment with aripiprazole, olanzapine, quetiapine, and risperidone in youth. Journal of Child and Adolescent Psychopharmacology. 2016;26:458–470.

Asai Y, et al. Machine learning-based prediction of digoxin toxicity in heart failure: a multicenter retrospective study. Biological and Pharmaceutical Bulletin. 2023;46:614–620.

Bae JH, et al. Machine learning for detection of safety signals from spontaneous reporting system data. Frontiers in Pharmacology. 2021;11:602365.

Blagus R, Lusa L. SMOTE for high-dimensional class-imbalanced data. BMC Bioinformatics. 2013;14:106.

Blake KV, et al. Comparison between paediatric and adult suspected adverse drug reactions reported to the European Medicines Agency. Paediatric Drugs. 2014;16:309–319.

Çorbacıoğlu ŞK, Aksel G. Receiver operating characteristic curve analysis in diagnostic accuracy studies. Turkish Journal of Emergency Medicine. 2023;23:195–198.

Chen C, et al. XGBoost-based machine learning test improves hemorrhage prediction among geriatric patients with rivaroxaban. BMC Geriatrics. 2023;23:418.

Choi H, et al. Machine learning model for predicting contrast-induced nephropathy. Internal Medicine. 2024;63:773–780.

Collins GS, et al. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD). Circulation. 2015;131:211–219.

Artificial intelligence and machine learning in software as a medical device. 2025.

Hu Q, et al. Predicting adverse drug event using machine learning based on electronic health records: systematic review and meta-analysis. Frontiers in Pharmacology. 2024;15:1497397.

GAMP 5 Guide. 2nd ed. 2025.

Jahid MJ, Ruan J. Ensemble approach for drug side effect prediction. IEEE BIBM Proceedings. 2013:440–445.

Jamal S, et al. Predicting neurological adverse drug reactions using machine learning. Scientific Reports. 2017;7:872.

Kim S, et al. PubChem substance and compound databases. Nucleic Acids Research. 2016;44:D1202–D1213.

Kuhn M, et al. The SIDER database of drugs and side effects. Nucleic Acids Research. 2016;44:D1075–D1079.

Lee SW, et al. Multi-center validation of machine learning model. NPJ Digital Medicine. 2022;5:91.

Lotfi Shahreza M, et al. Network-based approaches to drug repositioning. Briefings in Bioinformatics. 2018;19:878–892.

Ramspek CL, et al. External validation of prognostic models. Clinical Kidney Journal. 2021;14:49–58.

Rieder M. Adverse drug reactions in children. Journal of Pediatric Pharmacology and Therapeutics. 2019;24:4–9.

Sandhu S, et al. Integrating a machine learning system into clinical workflows. Journal of Medical Internet Research. 2020;22:e22421.

Snell KIE, et al. TRIPOD-SRMA checklist. BMJ. 2023;381:e073538.

Sun C, et al. Drug side-effect prediction based on comprehensive drug similarity. IFMCA Proceedings. 2017:171–178.

Ward IR, et al. Explainable artificial intelligence for pharmacovigilance. Computer Methods and Programs in Biomedicine. 2021;212:106415.

Wu L, et al. Hybrid deep forest-based method for predicting synergistic drug combinations. Cell Reports Methods. 2023;3:100411.

Yang J, et al. Machine learning generalizability across healthcare settings. NPJ Digital Medicine. 2022;5:69.

Zhang W, et al. Predicting drug side effects by multi-label learning and ensemble learning. BMC Bioinformatics. 2015;16:365.

Zhao J, et al. Predictive modeling of structured electronic health records. BMC Medical Informatics and Decision Making. 2015;15:S1.

Zhao X, et al. Similarity-based method for prediction of drug side effects. Mathematical Biosciences. 2018;306:136–144.

Zhou F, et al. Graph neural network-based subgraph analysis for predicting adverse drug events. Computers in Biology and Medicine. 2024;183:109282.

Zhu SY, et al. Interpretable machine learning model predicting immune checkpoint inhibitor-induced hypothyroidism. Cancer Science. 2024;115:3767–3775.