Table 2.
Do host characteristics predict bacterial, metabolite, and functional pathway abundances in the anal gland?
| Data type |
Distance type |
Age (yrs) |
Obesity (obese vs. not) |
Living environment (indoor vs. indoor outdoor) |
Diet (Dry food only vs. Other) |
Medical diagnosis (periodontitis) |
|---|---|---|---|---|---|---|
| Microbiome | Jaccard | 0.092 (p = 0.013) | 0.087 (p = 0.034) | 0.054 (p = 0.18) | 0.029 (p = 0.77) | 0.024 (p = 0.9) |
| Bray-Curtis | 0.12 (p = 0.003) | 0.029 (p = 0.74) | 0.043 (p = 0.37) | 0.042 (p = 0.43) | 0.034 (p = 0.57) | |
| Aitchison | 0.068 (p = 0.09) | 0.041 (p = 0.47) | 0.055 (p = 0.20) | 0.054 (p = 0.22) | 0.046 (p = 0.35) | |
| Metabolome (solid-phase) | Jaccard | 0.016 (p = 0.95) | 0.017 (p = 0.91) | 0.029 (p = 0.67) | 0.057 (p = 0.27) | 0.033 (p = 0.61) |
| Euclidean | 0.025 (p = 0.91) | 0.028 (p = 0.86) | 0.035 (p = 0.61) | 0.06 (p = 0.2) | 0.039 (p = 0.5) | |
| Metabolome (liguid-derivatization) | Jaccard | 0.051 (p = 0.33) | 0.045 (p = 0.39) | 0.057 (p = 0.26) | 0.019 (p = 0.88) | 0.035 (p = 0.57) |
| Euclidean | 0.044 (p = 0.41) | 0.039 (p = 0.50) | 0.061 (p = 0.17) | 0.024 (p = 0.86) | 0.037 (p = 0.54) | |
| COG functions | Jaccard | 0.049 (p = 0.24) | 0.049 (p = 0.29) | 0.12 (p = 0.025) | 0.026 (p = 0.7) | 0.021 (p = 0.79) |
| Bray-Curtis | 0.13 (p = 0.005) | 0.045 (p = 0.32) | 0.04 (p = 0.4) | 0.038 (p = 0.44) | 0.029 (p = 0.63) | |
| Aitchison | 0.085 (p = 0.01) | 0.045 (p = 0.32) | 0.05 (p = 0.23) | 0.041 (p = 0.43) | 0.037 (p = 0.56) | |
| KEGG functions | Jaccard | 0.045 (p = 0.31) | 0.045 (p = 0.33) | 0.11 (p = 0.039) | 0.022 (p = 0.8) | 0.019 (p = 0.86) |
| Bray-Curtis | 0.13 (p = 0.008) | 0.037 (p = 0.43) | 0.045 (p = 0.29) | 0.036 (p = 0.44) | 0.028 (p = 0.64) | |
| Aitchison | 0.092 (p = 0.01) | 0.041 (p = 0.46) | 0.05 (p = 0.21) | 0.042 (p = 0.41) | 0.035 (p = 0.61) | |
| Metagenome-assembled genomes (MAGs) | Jaccard | 0.051 (p = 0.23) | 0.040 (p = 0.58) | 0.068 (p = 0.04) | 0.045 (p = 0.41) | 0.045 (p = 0.4) |
| Bray-Curtis | 0.10 (p = 0.004) | 0.038 (p = 0.49) | 0.058 (p = 0.11) | 0.041 (p = 0.47) | 0.36 (p = 0.61) | |
| Aitchison | 0.085 (p = 0.014) | 0.027 (p = 0.91) | 0.045 (p = 0.43) | 0.036 (p = 0.7) | 0.034 (p = 0.75) |
Kraken2 Genus-level bacterial relative abundances were used to estimate microbiome Jaccard, Bray-Curtis or Aitchison dissimilarity distances. Metabolite GC-MS intensity values were normalized, log10-transformed and scaled to estimate Jaccard and Euclidean distances. Metagenome sequences mapping to putative gene or gene pathways were converted from Transcripts per Million to relative abundances to calculate Jaccard, Bray-Curtis or Aitchison matrices. The number of shotgun sequences mapping to MAGs was used to calculate relative abundances to construct the same three dissimilarity matrices. PERMANOVA models using 999 permutations correlated five host predictors with bacterial, metabolite, or functional beta-diversity. All predictors were evaluated simultaneously – in a way where the order of terms did not influence the statistical outcome. Significant p-values (a = 0.05) are bolded.