Graphene and metal-organic framework hybrids for high-performance sensors for lung cancer biomarker detection supported by machine learning augmentation

Abstract

Conventional diagnostic methods for lung cancer, based on breath analysis using gas chromatography and mass spectrometry, have limitations for fast screening due to their limited availability, operational complexity, and high cost. As potential replacement, among several low-cost and portable methods, chemoresistive sensors for the detection of volatile organic compounds (VOCs) that represent biomarkers of lung cancer were explored as promising solutions, which unfortunately still face challenges. To address the key problems of these sensors, such as low sensitivity, high response time, and poor selectivity, this study presents the design of new chemoresistive sensors based on hybridised porous zeolitic imidazolate (ZIF-8) based metal–organic frameworks (MOFs) and laser-scribed graphene (LSG) structures, inspired by the architecture of the human lung. The sensing performance of the fabricated ZIF-8@LSG hybrid sensors was characterised using four dominant VOC biomarkers, including acetone, ethanol, methanol, and formaldehyde, which are identified as metabolomic signatures in lung cancer patients’ exhaled breath. The results using simulated breath samples showed that the sensors exhibited excellent performance for a set of these biomarkers, including fast response (2–3 seconds), a wide detection range (0.8 ppm to 50 ppm), a low detection limit (0.8 ppm), and high selectivity, all obtained at room temperature. Intelligent machine learning (ML) recognition using the multilayer perceptron (MLP)-based classification algorithm was further employed to enhance the capability of these sensors, achieving an exceptional accuracy (approximately 96.5%) for the four targeted VOCs over the tested range (0.8–10 ppm). The developed hybridised nanomaterials, combined with the ML methodology, showcase robust identification of lung cancer biomarkers in simulated breath samples containing multiple biomarkers and a promising solution for their further improvements toward practical applications.