Assessment of cluster analysis of elastic light scatter profiles for the identification of foodborne bacteria

Abstract

Elastic Light Scatter (ELS) profiling is a novel approach for simultaneous detection and identification of bacteria cultured on solid agar media. The profiles comprise a range of different scatter features that can be used jointly or individually as a basis for comparison. We examined the utility of cluster analysis of ELS profiles for classification and identification of bacteria of relevance to foods. A total of 1562 colonies from 48 strains, representing 17 different species distributed among four genera, were examined. Each of three scatter-derived features (Zernike moments, pseudo-Zernike moments, proprietary Patsekin elements) were used individually and in combination for the cluster analysis. Of these, a combination of Patsekin elements and pseudo-Zernike moments yielded clusters that best reflected the known taxonomic relationships among the strains examined. Evidence of Genus- level markers of colony architecture was seen and there was a general agreement of clustering at the species level. Nonetheless, some individual colonies did not cluster with the majority of others from the same taxon, which could reflect an aberrant ELS phenotype, or known challenges in depicting strain relationships using cluster analytical methods. However, when compared with UMAP data processing, relationships between individual colonies were more easily discerned by inspecting the dendrogram. Cluster analysis of ELS profiles is a useful adjunctive tool for the classification and identification of bacteria and results may also be helpful in informing the development and improvement of other data analytical tools for ELS profile analysis.