A compact and reconfigurable terahertz (THz) bandpass filter based on a hybrid graphene–vanadium dioxide (VO2) architecture is presented for integrated filtering and biomolecular sensing applications. The proposed design employs a transmission-line–loaded stepped impedance resonator (SIR) configuration, achieving a stable bandpass response centred at 1.89 THz with an absolute bandwidth of 0.41 THz. Multi-modal tunability is realized through geometric variation, graphene chemical potential modulation (0.08–0.11 eV), and VO2 phase-transition-induced conductivity control, enabling frequency reconfiguration up to 1.95 THz. The structure is further utilized for refractive-index-based sensing, exhibiting a high sensitivity of 0.040 THz/RIU. The sensor is validated using biomolecular samples, specifically varying concentrations of haemoglobin, demonstrating distinct resonance shifts and effective concentration-level discrimination. In addition, the presence of a stable reference mode enables differential sensing, improving detection reliability. A machine learning-based regression framework is developed to model the relationship between refractive index and resonant frequency. Among the evaluated models, polynomial regression achieves the best performance with an R2 of 0.998, enabling accurate prediction and inverse estimation of analyte properties. The proposed approach provides a compact, tunable, and intelligent THz sensing platform with strong potential for real-time biomedical diagnostics, environmental monitoring, and advanced communication systems.