Machine learning-driven nanoparticle-enhanced paper chromogenic array sensor approach for detecting sub-lethally...

Detection of sub-lethally injured pathogens is critical for improving food safety, particularly given regulatory recommendations that microbiological method validation include 50-80% injured cells. In food matrices like low moisture foods (LMFs), injured cells can resuscitate and proliferate under favorable conditions, posing significant risks to public health. Furthermore, these cells may retain or even enhance virulence, underscoring the need for timely and accurate detection. This study developed a machine learning-driven nanoparticle-enhanced paper chromogenic array sensor (ML-NP-PCA) approach for detecting and differentiating injured and normal Salmonella in peanut butter (a type of high-fat LMFs) with background microflora (BK). Gold NP, silica NP, and zeolite NP were evaluated for enhancing PCA's performance. Among them, silica NP-PCA showed robust capability for identifying injured and normal Salmonella across a wide concentration range (∼1-6 log CFU/mL) in phosphate-buffered saline and was integrated in the ML-NP-PCA approach development. Results indicated that the ML-NP-PCA approach could accurately and continuously detect and distinguish between injured and normal Salmonella in peanut butter, even in the presence of BK, during a 48-h storage period at room temperature, with an accuracy of over 90%. Both injured and normal Salmonella could be detected as early as 1 h at ∼3-4 log CFU/g, with 92.0 ± 0.9% accuracy. These findings demonstrate the potential of the ML-NP-PCA approach as a non-destructive, enrichment-free, and rapid tool for continuous monitoring of injured Salmonella in foods. This approach also supports regulatory-aligned microbial testing and enhances food safety surveillance across the food supply chain.