Effective detection of indoor fungal contamination through the identification of volatile organic compounds using mass spectrometry and machine learning.

Indoor fungal contamination poses significant challenges to human health and indoor air quality. This study addresses an effective approach using mass spectrometry and machine learning to identify microbial volatile organic compounds (MVOCs) originated from indoor fungi. Three common indoor fungi, including Penicillium Chrysogenum, Cladosporium cladosporioides, and Aspergillus niger, were cultivated on various substrates, namely potato dextrose agar, wallpaper, and silicone. Solid-phase microextraction-gas chromatography-mass spectrometry (SPME-GC-MS) was used to analyze MVOCs together, along with the VOCs (namely, non-MVOCs) emitted by various indoor materials (wallpaper adhesives, diffusers, particle board, oriented strand board, medium-density fiberboard, bleach, print cartridges, and cosmetic creams). This study demonstrates the significant effectiveness of machine learning, particularly the random forest model, in accurately distinguishing MVOCs from non-MVOCs. Furthermore, specific VOCs such as benzocyclobutane, styrene, ethanol, benzene, and 2-butanone emerged as consistent indicators of fungal presence across different fungal species and substrates. A simplified random forest model incorporating these key VOCs achieved a high accuracy of 96.2%, highlighting their practical significance in fungal contamination detection. This integrated approach combining analytical chemistry and machine learning offers a promising strategy for the comprehensive and reliable assessment of indoor fungal contamination.