PharmaGuard: An AI-Driven Pharmacogenomic Risk Prediction System Using Explainable Clinical Decision Support

Pharmacogenomics plays a significant role in precision medicine by enabling personalized drug therapy based on an individual’s genetic profile. Variations in pharmacogenes such as CYP2D6, CYP2C19, CYP2C9, SLCO1B1, TPMT, and DPYD can significantly influence drug metabolism, efficacy, and toxicity, leading to adverse drug reactions (ADRs) or therapeutic failure if not properly identified. This paper presents PharmaGuard, an AI-driven pharmacogenomic risk prediction and clinical decision support system designed to analyze genomic Variant Call Format (VCF) files and provide explainable drug-response recommendations. The proposed framework integrates genomic variant parsing, pharmacogene mapping, CPIC-guideline-based interpretation, explainable artificial intelligence (XAI), and large language model (LLM)-based clinical explanation generation within a secure web-based platform. PharmaGuard classifies drug response outcomes into categories such as Safe, Adjust Dosage, Toxic, Ineffective, and Unknown, while generating interpretable recommendations for clinicians and researchers. The system supports multiple clinically relevant drug-gene interactions, including Codeine–CYP2D6, Clopidogrel–CYP2C19, Warfarin–CYP2C9, Simvastatin–SLCO1B1, Azathioprine–TPMT, and Fluorouracil–DPYD. Experimental evaluation demonstrates effective variant detection, real-time genomic analysis, and explainable clinical interpretation. The proposed framework contributes toward the development of intelligent, transparent, and scalable precision medicine systems capable of improving drug safety and personalized therapeutic