Compositional Data Analysis of Periodontal Disease Microbial Communities

Abstract

Periodontal disease (PD) is a chronic, progressive polymicrobial disease that induces a strong host immune response. Culture-independent methods, such as next-generation sequencing (NGS) of bacteria 16S amplicon and shotgun metagenomic libraries, have greatly expanded our understanding of PD biodiversity, identified novel PD microbial associations, and shown that PD biodiversity increases with pocket depth. NGS studies have also found PD communities to be highly host-specific in terms of both biodiversity and the response of microbial communities to periodontal treatment. As with most microbiome work, the majority of PD microbiome studies use standard data normalization procedures that do not account for the compositional nature of NGS microbiome data. Here, we apply recently developed compositional data analysis (CoDA) approaches and software tools to reanalyze multiomics (16S, metagenomics, and metabolomics) data generated from previously published periodontal disease studies. CoDA methods, such as centered log-ratio (clr) transformation, compensate for the compositional nature of these data, which can not only remove spurious correlations but also allows for the identification of novel associations between microbial features and disease conditions. We validated many of the studies' original findings, but also identified new features associated with periodontal disease, including the genera Schwartzia and Aerococcus and the cytokine C-reactive protein (CRP). Furthermore, our network analysis revealed a lower connectivity among taxa in deeper periodontal pockets, potentially indicative of a more "random" microbiome. Our findings illustrate the utility of CoDA techniques in multiomics compositional data analysis of the oral microbiome.

Keywords: C-reactive protein; CLR; compositional data analysis; microbiome; oral microbiome; periodontal disease.

FIGURE 1

Schematic of the analyses performed on the centered log-ratio (clr)-transformed periodontal treatment (PT) and sodium hypochlorite (SHT) data. Only untransformed data were inputted into selbal, and clr-transformed data were used to calculate Spearman’s correlations, non-metric multidimensional scaling (NMDS), and DIABLO analyses.

FIGURE 2

Non-metric multidimensional scaling (NMDS) ordination plots showing clustering of the samples. For the periodontal treatment (PT) study (A–F), columns correspond to dataset type: 16S, cytokines, and metagenomics are columns 1, 2, and 3, respectively (n = 60, 104, and 22). The top row (A–C) is colored by periodontal treatment and the second row (D–F) colored by overall response. For the sodium hypochlorite (SHT) study (G–I), columns correspond to dataset type: 16S, metabolomics, and metagenomics are columns 1, 2, and 3, respectively (n = 209, 153, and 24). Plots are colored by disease class.

FIGURE 3

Correlation structure between datasets as determined by the mixOmics DIABLO framework colored by pre- vs. posttreatment for the periodontal treatment (PT) study (A) bacterial 16S and cytokine datasets and (B) bacterial 16S, cytokine, and metagenomic datasets and also colored by whether the disease improved or worsened for (C) the bacterial 16S and cytokine datasets and (D) the bacterial 16S, cytokine, and metagenomic datasets. For the sodium hypochlorite (SHT) study, samples were colored by disease class for (E) the bacterial 16S and metabolic datasets and (F) the bacterial 16S, metabolites, and metagenomic datasets. Values indicate the between-dataset correlation structure. Ellipses indicate discrimination by the multiomics components between samples by condition.

FIGURE 4

Circos plots show 16S and other features with inter-omics links indicating positive or negative correlations with an R² greater than 0.5 between (A) the periodontal treatment (PT) study for the improved and worsened samples with the strongest correlation (0.6) for IL-6/IL-10 positively correlated with Schwartzia and Treponema and (B) the sodium hypochlorite (SHT) study for correlations with an R² greater than 0.7 for disease classes “A” and “C.” Bold labels were the strongest correlations.

FIGURE 5

Results of random forest Gini importance bar plots. Periodontal treatment (PT) dataset high/low periodontal disease (PD) using (A) 16S and (B) 16S and cytokines. Sodium hypochlorite (SHT) dataset A/C disease classes (C) 16S and (D) 16S and metabolites. Larger Gini importance indicates that features can resolve more nodes in decision trees with more confidence.

FIGURE 6

Network visualizations. Network plots derived from the periodontal treatment (PT) study patients whose oral status was designated as (A) gingivitis, (B) moderate, (C) shallow, or (D) deep. For the sodium hypochlorite (SHT) study, network plots for biofilms from patients with (E) class “A” or (F) class “C.” Node size corresponds to eigen centrality. Dashed lines represent negative correlations and solid lines represent positive correlations.