Like Will to Like: Abundances of Closely Related Species Can Predict Susceptibility to Intestinal Colonization by Pathogenic and Commensal Bacteria (original) (raw)
Figure 6
In CON mice, related bacterial lineages are preferentially observed together (quantitative co-occurrence).
For all possible pairs of detected OTUs (i.e. present in at least 2/3 of the analyzed mice; CD = 0.2; each dot in the graph represents an OTU-pair), abundance correlations (y-axis, Pearson) were computed from abundance measurements in 9 distinct CON mice, and plotted against the molecular divergence between their representative 16S sequences (x-axis). The latter distances between representative sequences were computed using sequence identities as defined by the complete multiple alignment of all reads and all reference sequences. For hypothesis testing, we compared the data distribution (red) to a matched random distribution of OTU abundances generated by shuffling non-null OTU abundances between all OTUs (blue). Running medians are represented in the corresponding color. The Pearson correlation coefficient is −0.248 (p-value = 7.10e-18) for the actual data, and −0.017 (p-value = 0.563) for the randomized data. We compared the deviation of the actual data on the y-axis (Pearson correlation) from the distribution of the randomized data using the Kolmogorov-Smirnov test. Both two-sided and one-sided hypotheses (greater and less) were tested for each bin of 0.1 on the x-axis (0.0-0.1; 0.1–0.1; 0.2–0.3; 0.3–0.4; 0.4–0.5). Results are indicated in boxes in the upper part of the graph. Pv = P-value; ‘KS two-sided pv’ indicates whether there is a significant difference between the distribution of the data (red dots) and the distribution of the randomized data (blue dots); ‘KS greater pv’ indicates whether the Pearson correlation coefficient of the data (red dots) is significantly higher compared to the random background in a given bin (blue dots). ‘KS less pv’ indicates whether the Pearson correlation coefficient of the data (red dots) is significantly lower compared to the random background in (blue dots) in a given bin. The histogram on the right side of the graph represents the cumulative frequencies of the binned Pearson correlation coefficient data.