SERS-AI Based\nDetection and Bioanalysis of Malodorous\nComponents in Kitchen Waste

The prevention and control of odor gas generated from\nkitchen waste\nare significant missions in research on environmental pollution. Because\nof the high complexity and variability of kitchen waste, the development\nof a suitable technique with high sensitivity for the accurate detection\nof odor gas is an urgent and core task in this frontier field. Here,\na technique combining surface-enhanced Raman spectroscopy (SERS) and\nartificial intelligence (AI) is explored for detecting malodorous\ncomponents in the leachate of kitchen waste. Initially, 1706 SERS\nspectra were collected from synthetic kitchen waste under various\nfermentation parameters. Several AI algorithms were used to classify\nthree levels of odor intensity based on SERS spectra, among which\nthe Random Forest Classifier algorithm model showed a high prediction\naccuracy of 86.5%. Then, by integrating Raman data, the AI algorithm\nmodel identified hydrogen sulfide (H<sub>2</sub>S) and ammonia (NH<sub>3</sub>) as the dominant malodorous components in the odor gas. Finally,\nthe structural characteristics of the microbial communities are investigated.\nWith the help of Raman’s intensities of malodorous components,\nmany more insights into microorganisms in the fermentation process\nare revealed, which has important research value in the prevention\nand controlling of odor gas generated from kitchen waste. Furthermore,\nthe microbial metabolic pathways of sulfur and nitrogen are discussed\nhere. This SERS-AI-based novel technique not only has a broad potential\nfor odor pollution but also could be applied to another complicated\nbiochemical system with functional bacteria.