Fig. 10

Analysis of stool microbiomes of colorectal cancer. a Data ordination through algorithm-based (PCA and t-SNE), model-based (ZIFA, PPCA-NB, mbDenoise-zinb), and denoising methods (PPCA-NB, mbDenoise-zinb and SAVER) by applying PCA and t-SNE to the denoised data on the third dataset in Table 1. Inputs of PCA and t-SNE were log-transformed, and the empty was due to an exception in ZIFA. Beta diversity was assessed using PERMANOVA. b Alpha diversity analysis on the third dataset in Table 1. Included methods for composition estimation were zr, svt, pmr, dmm, empirical Bayes estimate by PPCA-NB and by mbDenoise-zinb, and zr using the denoised data from SAVER (SAVER_zr). Significance was calculated using the Wilcoxon test