Graphene FET biochip on PCB reinforced by machine learning for ultrasensitive parallel detection of multiple antibiotics in water.

Antibiotics like Ciprofloxacin (Cfx), tetracycline (Tet) and Tobramycin (Tob) are commonly used against a broad-spectrum of bacterial infection. Recent surge in their uptake through the presence of their residues in environmental water has been linked to increased antibiotic resistance. Conventional methods for antibiotic monitoring by gold standards like LC-MS though sensitive and reliable, are expensive, requires dedicated equipment and complex sample processing steps. In this context, nanoscale field-effect transistors (FETs) present significant advantages of rapid measurement and ultra-high sensitivity but the device-device variations in the transfer characteristics originating from the inherent fluctuations in fabrication protocol of 2D materials, lead to stochasticity in bioreceptor orientation and binding densities which limits their potential for ultrasensitive and reliable detection of multiple antibiotics in river water. Here, we introduce a distinctive approach for few femtomolar detection of Cfx, Tet and Tob simultaneously in river water by developing thermally reduced graphene oxide (TRGO) FET array on printed circuit board utilizing copper plated electrodes where multiple features extracted from sensor transfer characteristics are processed by machine learning models, trained with moderate calibration dataset. The demonstrated methodology detects 1 fM concentration of Cfx, Tet and Tob with satisfactory accuracy within 20 min, using XGBoost model. The achieved detection limit is three and two orders of magnitude lower than previous reports of multiple and single antibiotic detection respectively. The TRGO FET sensor array interfaced with an electronic readout imparts capability to track the concentration of antibiotic contaminants in various water sources and adopt necessary measures for safe drinking water.