MODELLING AND MITIGATING ANTIMICROBIAL RESISTANCE (AMR) THROUGH DATA-DRIVEN SURVEILLANCE, AI-POWERED DRUG DISCOVERY, AND PUBLIC HEALTH INTERVENTION DESIGN

Antimicrobial resistance (AMR) has become arguably the greatest threat to global health, eroding decades of progress in the control of infectious disease and straining public health infrastructure, especially in low- and middle-income countries (LMICs). The accelerating trend of multidrug-resistant microorganisms, driven by the misuse of antibiotics, poor surveillance, and constrained drug development, calls for a combined, evidence-based approach that unites technology and public health interventions. The objective of this research is to design and validate a comprehensive model for modelling and managing AMR using quantitative techniques such as epidemiological surveillance analysis, artificial intelligence-based antimicrobial compound discovery, and simulation model-based public program design. Retrospective multi-phased quantitative research was used. AMR surveillance data (2010–2024) were retrieved from GLASS, NCDC, and Nigeria hospitals to determine resistance prevalence, risk factors, and geospatial distribution. Statistical and machine learning algorithms, such as logistic regression and random forest, were used to make predictions of multidrug resistance. Molecular descriptors of 10,000 molecules from ZINC15 and ChEMBL were explored using AI models (XGBoost, GNNs) to make predictions of antimicrobial activity for drug discovery. Molecular docking and ADMET profiling were conducted on high-confidence compounds. Agent-based and system dynamics models then simulated the effects of diverse public health interventions on a 10-year time scale. Surveillance data analysis showed a 42.3% prevalence of AMR, and E. coli and K. pneumoniae were the strongest resistant microorganisms. Prior antibiotic exposure history (OR = 3.42) and hospital-acquired infection (OR = 2.91) were strong predictors. Excellent accuracy was predicted by machine learning models (XGBoost AUC = 0.93) and identified 25 new compounds with high binding propensity (up to –10.2 kcal/mol) against resistance proteins. Simulation findings indicated that coordination among antibiotic stewardship, public education, and restricting the sales of OTC products would reduce AMR rates by 55% in 10 years and save about $11.3 million in healthcare every year. This study shows that the combination of AI, large data analysis, and predictive simulation offers a powerful framework for AMR modelling and intervention planning. The results show how multisectoral policies with combined technological advances and behaviour-specific public health interventions are needed. Implementation of such a combined model can safely limit AMR trends, advise national antimicrobial policy, and hasten the development of new treatment avenues in Nigeria and other comparable LMIC settings.