Fig. 5

Dominant microorganisms among the SAMR1_PBS group, SAMP8_PBS group, and SAMP8_MSC group (n = 7 per group). A, C, E LEfSe clustering trees comparing differential species between SAMR1_PBS and SAMP8_PBS (A), SAMP8_PBS and SAMP8_MSC (C), and SAMR1_PBS and SAMP8_MSC (E) groups pairwise. As depicted in LEfSe clustering tree, different colors represent different groups, with color_coded nodes indicating microbial taxa that play a significant role within each group. Each colored dot represents a biomarker, with biomarker names provided in the legend at the bottom right corner. Yellow nodes denote microbial taxa that do not play a significant role across different groups. Taxonomic levels from inner to outer circles include phylum, class, order, family, and genus. LEfSe is a software used for discovering high_dimensional biomarkers and revealing genomic features. It encompasses genes, metabolism, and taxonomy to discriminate between two or more biological conditions (or groups), emphasizing statistical significance and biological relevance, enabling researchers to identify differentiating features and associated taxonomic categories. It employs biological statistical differences for robust identification and conducts additional tests to assess whether these differences conform to expected biological behavior. Software used: LEfSe (https://huttenhower.sph.harvard.edu/galaxy/). B, D, F LDA plots comparing differential species between SAMR1_PBS and SAMP8_PBS (B), SAMP8_PBS and SAMP8_MSC (D), and SAMR1_PBS and SAMP8_MSC (F) groups pairwise, displaying LDA scores corresponding to species with p < 0.05. LDA plots illustrating significant microbial taxa within different groups. Colors represent microbial taxa significantly influencing different groups. The plot primarily displays biomarkers with LDA scores greater than the preset value (2.0). The color of the bars represents the respective groups, while the length represents the LDA score, indicating the magnitude of the significant differences between groups. LDA is a classic and popular algorithm in the field of machine learning and data mining, serving as a supervised dimensionality reduction technique, where each sample in the dataset has a categorical output. This differs from PCA, which is an unsupervised dimensionality reduction technique that does not consider sample category outputs. Compared to PCA analysis, LDA algorithm effectively utilizes sample grouping information, resulting in more reliable outcomes. Software used: LEfSe (https://huttenhower.sph.harvard.edu/galaxy/)