Artificial intelligence and Machine learning in pharmacy and pharmaceutical technology

Abstract

Artificial intelligence (AI) and Machine learning (ML) are transforming the pharmaceutical lifecycle—from target identification and lead optimization to clinical development, manufacturing, supply chain orchestration, and real-world pharmacovigilance. This review synthesizes recent advances (2018–2025) across core methodological families (supervised, unsupervised, reinforcement, and deep learning) and maps them to high-impact use cases in pharmacy and pharmaceutical technology. We systematically screen academic and industry literature using PRISMA guidelines, extracting evidence on model classes, datasets, performance metrics, validation strategies, and regulatory considerations. Key progress includes foundation and multimodal models for de novo molecular design, active learning for assay triage, graph neural networks and transformer architectures for property prediction and ADMET profiling, and digital twins and predictive control for continuous manufacturing. In clinical and post-marketing contexts, federated learning enhances privacy-preserving trial analytics, while NLP models accelerate signal detection in safety reports and EHRs. Despite clear gains in speed, cost, and decision quality, translation hurdles persist: data heterogeneity and bias, limited external validity, explainability gaps in high-stakes decisions, GMP/21 CFR Part 11/GxP-aligned validation, and evolving regulatory expectations for Software as a Medical Device and AI-as-a-Service. We outline best practices for reproducibility, model risk management, and human-AI governance, and identify research frontiers in causal ML, uncertainty quantification, hybrid physics-ML models, and responsible deployment. The review provides a consolidated roadmap for researchers, practitioners, and regulators to harness AI/ML for more efficient, safe, and equitable pharmaceutical innovation.