Generative artificial intelligence in health sciences: A comprehensive review of emerging applications, recent advancements, challenges and future perspectives

Generative artificial intelligence (GenAI) is becoming an important computational approach in health sciences because it can create new molecular structures, protein sequences, scientific text, images and synthetic data from patterns learned in large datasets. Unlike conventional predictive artificial intelligence, which mainly estimates an outcome from existing inputs, generative systems can propose previously unobserved candidates and produce structured content for further testing. This review examines the principles, emerging applications, recent advances, limitations and future directions of GenAI across the pharmaceutical lifecycle. Major application areas include de novo small-molecule design, lead optimisation, protein and peptide engineering, target and mechanism research, drug repurposing, formulation development, preclinical assessment, clinical-trial operations, pharmaceutical manufacturing, quality systems and pharmacovigilance. Developments in variational autoencoders, generative adversarial networks, transformers, large language models and diffusion models have expanded the range of molecules and biological structures that can be explored computationally. Protein-structure tools such as AlphaFold 3 and diffusion-based design systems illustrate the growing integration of generative modelling with structural biology. Nevertheless, generated candidates are hypotheses rather than medicines: they require chemical synthesis, analytical characterisation, biological testing, toxicological assessment and clinical evaluation. Data quality, model hallucination, limited external validation, bias, interpretability, intellectual-property uncertainty, privacy and regulatory accountability remain major concerns. The review concludes that the most credible near-term model is humansupervised, experimentally grounded use of GenAI within a reproducible design–make–test– analyse workflow. GenAI is likely to strengthen pharmaceutical research productivity and broaden exploration of chemical and biological space, but it cannot replace experimental science, professional judgement or regulatory standards.