Design and Analysis of a Highly Sensitive Terahertz Biosensor Using Graphene Metasurfaces and Surface Plasmon Resonance for Protein Detection with AI-Assisted Locally Weighted Linear Regression for Behavior Prediction

Protein detection is essential across diverse biomedical and biochemical disciplines, including disease diagnostics, pharmaceutical research, and environmental monitoring. Precise and efficient detection methodologies are critical for explaining complex biological processes and identifying biomarkers associated with various health conditions. This study presents the development and evaluation of an advanced terahertz biosensor designed for protein detection, utilizing graphene metasurfaces combined with surface plasmon resonance (SPR) to achieve high sensitivity. The sensor design incorporates a circular ring resonator and rectangular resonators on a silica substrate, with a graphene layer serving as the active sensing element. Finite element simulations are conducted to optimize the sensor’s geometric and material parameters. The optimized sensor design demonstrates impressive performance characteristics, achieving an optimal sensitivity of 508 GHzRIU⁻¹ and a quality factor of 5.089. Further analysis reveals a high figure of merit (11.293 RIU⁻¹), low detection limit (0.607), and strong signal-to-noise ratio (0.022). Additionally, the sensor exhibits potential for use in 2-bit encoding applications. A locally weighted linear regression model is leveraged to analyze and predict the sensor’s performance across various parameter combinations, achieving R² scores of up to 100%. The proposed biosensor shows promise for diverse applications, including enhancing medical diagnostics, ensuring food safety, and monitoring environmental factors.