Despite significant advances in the diagnostic techniques used in microbiology, precise and rapid detection of pathogens remains a challenge. This is particularly concerning in food microbiology, where the detection of, and timely interventions for, spoilage and pathogenic microbes are critical in maintaining food quality and protecting public health (Ferone et al., 2020; White et al., 2023). Notably, misidentification of pathogens can lead to severe consequences and widespread outbreaks (El-Sayed et al., 2022; Maeda et al., 2017) as evident by global foodborne illnesses, with an annual estimate of two billion cases and one million deaths due to such infections being made (Kirk et al., 2015).

While traditional culture-based methods for microbial detection are well-established, they remain time-consuming and labor-intensive (E. Bae et al., 2012a; Ramadan, 2022). In contrast, culture-independent methods, like polymerase chain reaction (PCR) and DNA sequencing, have been prevalent because they are effective, rapid, sensitive, and have high specificity, but do not allow for detailed and subsequent analysis of isolates for thorough characterization (Arai et al., 2020; Lauri and Mariani, 2009). Effective outbreak control requires rapid methods that can accurately detect and differentiate pathogenic from non-pathogenic bacteria while preserving viable samples for subsequent epidemiological investigation (On et al., 2021b; Tang et al., 2019).

Elastic Light Scatter (ELS) analysis allows for the interrogation of distinct colonies using a laser beam, generating composite forward scatter patterns that act as species-specific signatures based on subtle structural differences (Bae et al., 2007; Banada et al., 2007; Bayraktar et al., 2006). Earlier studies have demonstrated that ELS possesses the capability for rapid, label-free, non-destructive detection and identification of foodborne pathogens, directly upon culture, enabling further subtyping of isolates for epidemiological analysis. This is a unique attribute of the method, making it a valuable adjunct to other genetic and phenotypic approaches. It has been used successfully to discriminate various bacteria of importance to food and food safety, including Vibrio, Campylobacter, Yersinia, Listeria, and Arcobacter species (Banada et al., 2009; He et al., 2015; Huff et al., 2012; On et al., 2021a, On et al., 2021b; Patsekin et al., 2019). The ELS approach has also demonstrated the capacity to detect previously uncharacterized species, underscoring its potential to identify novel and emerging pathogenic taxa not represented in the existing pattern database (Rajwa et al., 2010). However, in conventional identification scenarios (typical of many diagnostic systems) the accurate identification of microbial colonies on agar media depends on the presence of corresponding patterns in the training dataset. If the system is not explicitly designed to detect anomalies or emerging patterns, any isolate representing a taxon absent from the reference database will inevitably be misclassified. This was illustrated in a study of ELS analysis for the detection of foodborne pathogenic Yersiniae (On et al., 2021a). Here, 88 % of strains identified by ELS as belonging to Yersinia species were correctly identified. However, there were instances where strains were misidentified because they belonged to groups that were not included in the original database derived from pork meat studies.

Our study aimed to assess the utility of cluster analysis in identifying overlaps and similarities among taxa, using features (individually, and in combination) extracted via the ELS process. This approach has only been applied using one such feature previously (Zernike moments), to just 17 strains representing six Listeria species (Bayraktar et al., 2006). Here, we analyzed the spectral characteristics of multiple individual colonies from 17 different species of four genera for this purpose. The results of cluster analysis were compared with that of a non-hierarchical clustering approach, Uniform Manifold Approximation and Projection (UMAP), which has been found useful for determining interstrain relations among winemaking and brewing yeasts characterized by another high-resolution phenotyping method, Matrix-Assisted Laser Desorption-Ionisation Time-Of-Flight Mass Spectrometry (Zhang et al., 2022).