Background: The integration of Drug Decision Support Systems (DDSS) is a foundational element of modern medical practice designed to mitigate the global burden of adverse drug events. Driven by rapid advances in genomics, artificial intelligence (AI), and computational modeling, DDSS architectures have transitioned from primitive, rule-based engines into adaptive, data-driven frameworks engineered to support precision medicine.

Objective: This scoping review systematically synthesizes the methodological and architectural milestones of this technological transition, mapping the evolution from deterministic rules to advanced probabilistic and neural hybrid systems.

Methodology: Surveying core literature sources across PubMed, Scopus, and IEEE Xplore for the period spanning January 2012 to March 2024, this review systematically appraised and categorized the computational architectures of relevant peer-reviewed studies.

Results: The review highlights a structural shift toward probabilistic inference methods, evaluating Viterbi decoding to reduce noise in clinical sequence data, alongside deep learning models like Graph Neural Networks (GNNs) that refine drug-drug interaction (DDI) predictions. However, primary findings indicate that widespread clinical adoption is stalled by a pervasive reliance on retrospective simulation studies rather than prospective real-world trials, severe multi-omics data heterogeneity, and stark equity gaps due to minority genome underrepresentation in training sets.

Conclusion: Ultimately, realizing the full potential of personalized pharmacotherapy relies on resolving these structural deficits through federated learning infrastructures and transparent algorithmic governance. Achieving this clinical integration requires the deployment of highly interpretable, Explainable Artificial Intelligence (XAI) models capable of seamless execution within existing electronic health records.