Development and Validation of a Computational Predictive Model for Antimicrobial Resistance Gene Proliferation in Nigerian Agricultural Wastewater Systems

Agricultural wastewater from large-scale farms in Nigeria carries antibiotic residues, resistant bacteria, and resistance genes that enter receiving water bodies, yet computational models capable of predicting resistance gene proliferation dynamics under Nigerian agricultural wastewater conditions have not been developed. This dissertation develops and validates a computational predictive model for antimicrobial resistance gene (ARG) proliferation in Nigerian agricultural wastewater systems, integrating microbiological surveillance data, hydrological parameters, and machine learning into an original predictive modelling framework. Empirical data were collected from agricultural wastewater systems in Ogun (South West), Benue (North Central), and Rivers (South South) States over 18 months, measuring ARG concentrations by quantitative PCR targeting blaESBL, mcr-1, vanA, tetM, and sul1 genes alongside physicochemical and hydrological parameters. A hybrid modelling approach integrating mechanistic compartmental differential equations with a random forest machine learning layer was developed and trained on the empirical dataset. Cross-validation across held-out temporal datasets yielded a mean absolute prediction error of 12 percent for ARG concentration at downstream sampling points. The model identifies temperature, organic carbon loading, and upstream antibiotic application timing as the three strongest ARG proliferation predictors in Nigerian agricultural contexts. An open-access web-based model deployment tool was developed for use by Nigerian environmental health agencies. The dissertation makes original contributions to computational environmental microbiology and provides a regulatory decision support tool for ARG monitoring in Nigerian agricultural systems. Keywords: antimicrobial resistance genes, computational model, agricultural wastewater, Nigeria, predictive modelling