High-Sensitivity SPR Biosensor for Dual Detection of Glucose and Tuberculosis Biomarkers with XGBoost-Assisted Predictive Modelling

Tuberculosis (TB) and diabetes mellitus are widespread, with chronic hyperglycemia increasing TB susceptibility. Conventional diagnostics-sputum microscopy, culture assays, nucleic acid amplification, and blood glucose tests-require laboratory infrastructure, invasive sampling, or long processing times. Surface plasmon resonance (SPR) biosensors offer label-free detection by measuring refractive index changes at a metal-dielectric interface. This study presents a five-layer Kretschmann SPR sensor (BK7 prism/Ag/graphene/WS 2 /biosample) for glucose and TB biomarker detection. Optical behavior was analyzed via the Transfer Matrix Method and finite element simulations in COMSOL Multiphysics, with silver permittivity modeled by Drude-Lorentz and graphene conductivity by Kubo formalism. Layer thicknesses were optimized (Ag: 20-55 nm, graphene: 0.2-3.0 nm, WS 2 : 0.3-1.7 nm), with the configuration 30 nm Ag, 1.0 nm graphene, and 0.5 nm WS 2 yielding the strongest resonance (reflectance minimum 0.154%, angular sensitivity 400°/RIU, quality factor 460 RIU -1 ). The sensor exhibited linear resonance shifts with glucose (R 2 = 0.965, slope 170.4/RIU) and TB biomarkers (R 2 = 0.998, slope 163.0/RIU). Data from thickness optimization were used to train an XGBoost regression model, predicting Ag-thickness-dependent responses with R 2 = 0.9952 and RMSE = 0.0125, replicating simulation trends at far lower computational cost. These results demonstrate that the Ag/graphene/WS 2 multilayer enhances SPR sensitivity, providing an efficient platform for label-free, multi-analyte biochemical detection via angular interrogation.