Insights into the Complexities of Pharmacotherapy Parameters in Artificial Intelligence Models for Drug Selection, Precision Personalised Medicine and Optimal Therapeutic Outcomes

Background: Artificial intelligence (AI) is transforming pharmacology and pharmacotherapy by enabling the integration of large, heterogeneous datasets to support precision and personalised medicine. However, the reliability of AI-assisted therapeutic decision-making depends fundamentally on the selection, quality, and interpretation of pharmacological and clinical input features or parameters. Current AI models frequently overlook the multidimensional complexity of drug- and patient-specific factors, limiting their clinical applicability and generalisability. Methods: This narrative review was prepared based on 45 years of experimental work and relevant published research in drug design, development, and clinical experience in iron chelation therapy and personalised treatment approaches. Additional literature was identified through targeted searches of PubMed and Scopus using terms related to AI, machine learning, pharmacology, pharmacotherapy, and personalised medicine. Key pharmacological and clinical input features relevant to AI-assisted personalised drug selection include physicochemical drug properties, absorption, distribution, metabolism, excretion and toxicity characteristics, route of administration, drug interactions, pharmacokinetics, pharmacodynamics, multi-omics data, therapeutic efficacy, diagnostic biomarkers, statistical validation, governance, and explainable AI. A conceptual framework for AI-assisted personalised drug selection is also proposed as an example. Results: It is suggested that reliable AI-assisted pharmacotherapy requires the integration of diverse, interdependent datasets and parameters describing drug characteristics, patient variability, clinical outcomes, and real-world evidence. The incorporation of pharmacogenomics, electronic health records, diagnostic imaging and profiling, and validated computational descriptors can improve prediction of drug efficacy, toxicity, interactions, and therapeutic response. Furthermore, robust model validation, data governance, cybersecurity, and continuous monitoring are identified as essential prerequisites for safe clinical implementation. The proposed conceptual framework illustrates how clinical admissibility filtering, model-based ranking, local attribution, and clinically supervised decision support may enhance personalised drug selection while maintaining human oversight. Conclusions: Artificial intelligence has considerable potential to improve drug selection and personalised pharmacotherapy. However, its success depends on comprehensive integration of pharmacological knowledge with high-quality clinical data, rigorous validation, and responsible governance. The multidimensional framework presented here provides a foundation for the potential development of clinically interpretable, reliable, and patient-centred AI systems capable of supporting safer and more effective precision medicine.