Fig. 2: Tissue-resident microbiome in different anatomical locations of the colon and rectum.
From: Tissue-resident microbiota impacts colorectal cancer progression and prognosis
a Comparison of prokaryotic reads number between left- and right-sided tissue samples. Two-sided P-values were calculated with Wilcoxon rank-sum test. b Genus level microbial diversity in left- vs. right-sided tissue samples. Two-sided P-values were calculated with MaAsLin2, with adjustment for age, sex, and prokaryotic read count. For a and b: UU tumors: n (left)=386; n (right)=389; for UU NATs: n (left)=145, n (right)=168. For UM: n (left)=73, n (right)=89. c Enrichment of tissue-resident species between anatomical locations. Species displaying consistent differences between anatomical locations in both tumors (UU_T, n = 775; UM_T, n = 162) and NATs (UU_N, n = 313; UM_N, n = 162). MaAsLin2, BH-adjusted P < 0.05 (two-sided test). Green, right-side enriched specie; yellow, left-side enriched species. Colored outer bars indicate the corresponding phyla of presented species. d Genus-level microbial diversity in paired tumor-NAT samples across anatomical locations: all pairs (UU, n = 313; UM, n = 162), right-sided pairs (UU, n = 168; UM, n = 89), and left-sided pairs (UU, n = 145; UM, n = 73). Two-sided P-values were calculated with MaAsLin2. e Enrichment of tissue-resident species displaying significant and consistent differences between tumors and paired NATs in UU (168 UU_R and 145 UU_L) and UM (89 UM_R and 73 UM_L). MaAsLin2 (two-sided test), BH-adjusted P < 0.05 for UU, and P < 0.05 for UM (two-sided test). Blue, NAT-enriched; red, tumor-enriched. Colored outer bars indicate location-enriched species (yellow, left-sided; green, right-sided). f Two-sided 95% confidence interval for the area under the ROC curve (AUC) showing the ability of tissue-resident microbiota in distinguishing right from left-sided samples. Microbial based random forest (RF) models were generated using UU tumors (n = 775) with 5-fold cross validation (CV), and tested on UU_N (NATs, n = 313), UM_T (tumors, n = 162) and UM_N (NATs, n = 162). g Performance of RF models in distinguishing tumors from paired NATs. The training model was established on paired tissues of UU cohort (n = 313 pairs, 5-fold CV), and the testing set was 162 paired UM tissues. For the boxplots, boxes represent the interquartile range (IQR, Q1–Q3), the central line represents the median, and whiskers extend to 1.5 times of IQR.
