Page 88 - Read Online

Kamilari et al. Microbiome Res Rep 2025;4:3 https://dx.doi.org/10.20517/mrr.2024.47 Page 7 of 15

Figure 2. Microbial signatures discovery using the LEfSe algorithm, showing (A) bacterial taxa with statistically significant over-
representation among different areas; (B) fungal taxa with statistically significant over-representation among different areas; (C)
bacterial taxa with statistically significant over-representation between traditionally and industrially produced sausages; (D) fungal taxa
with statistically significant over-representation between traditionally and industrially produced sausages, based on a non-parametric
factorial KW sum–rank test, an (unpaired) Wilcoxon rank–sum test, and LDA. The phylogenetic trees map the taxonomic variances of
the identified microbial signatures from class (external part of the circle) to species level (internal part of the circle), combined with a list
of taxa with significantly increased representation in different sausage groups. KW: Kruskal–Wallis; LDA: linear discriminant analysis.

Differentiation of sausages based on origin and manufacturing conditions using random forest
algorithm
The random forest algorithm was trained to predict, based on the 16S rRNA gene abundance data, the
different origins of the analyzed sausages. The results revealed that sausages from different areas could be
separated with reliable predictive capability [area under the curve (AUC) = 0.99 regarding bacteria
(Figure 3A)]. This indicates that the algorithm could predict different sausages’ origins with great reliability.
The key predictive features that could differentiate sausages of different origins were similar to the microbial
signatures that were identified using the LefSe algorithm. Specifically, the taxa Lactobacillus and S. succinus
characterized traditional Pitsilia sausages, and Leuconostoc was typical of sausages from Nicosia. However,
Mytilini sausages were mostly characterized by the genera Streptococcus and Psychrobacter.

Cypriot traditional sausages were separated from Cypriot industrial sausages with high predictive accuracy
(average AUC = 0.97 for both bacteria and fungi, Figure 3B and C, respectively). Industrial sausages were
differentiated by the elevated presence of the bacterial taxa Erysipelotrichaceae, Parabacteroides,
Clostridium, Ruminococcus, Leuconostoc, Acinetobacter, Brochothrix, Pseudomonas, and Enterobacteriaceae,
whereas traditional sausages by the family Lactobacillaceae. Regarding the fungal diversity, industrial
sausages were distinguished by increased relative representation of Mucor flavus, Trichosporon,