Machine learning-integrated metal-organic frameworks/multi-walled carbon nanotubes sensor array for real-time detection of fungal VOCs.

Biological hazards posed by indoor fungi present significant challenges due to their hidden growth and the limitations of conventional detection methods. As an alternative, microbial volatile organic compounds (MVOCs) offer a non-invasive proxy for early-stage fungal contamination. This study presents a multi-modal gas sensor platform integrating metal-organic framework (MOF)/multi-walled carbon nanotube (MWCNT) composites with machine learning algorithms for real-time MVOC detection. Three MOF/MWCNT composites-ZIF-8/MWCNT, Cu-BTC/MWCNT, and UiO-66/MWCNT-were coated on three sensor types: quartz crystal microbalance (QCM), resistive sensor (RS), and electrochemical impedance spectroscopy (EIS). The sensor array was evaluated through exposure to single compounds, gas mixtures, and headspace emissions from common indoor fungi (i.e., Aspergillus, Cladosporium, and Penicillium). Ethanol, 2-butanone, and benzene were selected as test compounds. For single-gas detection, the random forest classification model achieved 95 % accuracy. For gas mixtures, the CatBoost regression model yielded R² values of 0.59 (ethanol), 0.93 (2-butanone), and 0.79 (benzene). Headspace sampling from fungal cultures enabled accurate classification of fungal emission versus control samples, achieving a classification accuracy of 0.963 using the Catboost model. These results confirm the feasibility of indirect, data-driven fungal detection using gas-phase signals. This work demonstrates a proof-of-concept sensing platform that holds promise for scalable and adaptive applications in indoor air quality monitoring. Future efforts will focus on scaling up MOF-composite synthesis and developing portable devices for early warning of microbial contamination in high-risk environments.