Fig. 4: Validation of L1- and L5-specific correlations in newly collected clinical CRC samples with FL-typing.

A L1 levels in faecal samples compared between CRC patients (n = 60) and healthy controls (n = 53). Mann‒Whitney test. B L1 levels were compared between tumour and matched normal tissues from 100 CRC patients. The p value of the Mann‒Whitney test is shown, whereas that of the Wilcoxon matched‒pairs signed‒rank test is 1.86e-11. C L5 levels in CRC tumours compared between lymphovascular invasion (LVI)-positive (n = 33) and LVI-negative tumours (n = 67). Mann‒Whitney test. Medians and interquartile ranges are shown. If not detected, −ΔCt values of -35 and -25 were manually defined for the faecal and tissue samples, respectively. D ROC curves of L1, ‘F. nucleatum’ and Fn-F/R for the previously and newly collected faecal samples (cohorts F1 and F2, respectively). ROC curves were compared with the DeLong method, Benjamini‒Hochberg correction was additionally applied, and adjusted p values are shown. E PR curves of L1, ‘F. nucleatum’ and Fn-F/R for cohorts F1 and F2. PR curves were compared with the bootstrap method followed by Benjamini‒Hochberg correction, and adjusted p values are shown. F ROC curves of L1 and ‘F. nucleatum’ for eight published faecal metagenomic datasets across diverse populations. L1 and ‘F. nucleatum’ abundance was represented by the FPKM values of their specific makers L1_746 and Fn_SR, respectively. The ROC curves were compared with the DeLong method. The corresponding PR curves are provided in Supplementary Fig. 7. For (D, E), ROC comparison with the bootstrap method was also conducted and the p or adjusted p values are listed in Supplementary Table 4. All AUCs are reported with their 95% confidence intervals in the parentheses.