paperKB
coga / coga-kb
Help
Sign in

Chunk #15 — Results — scCODA scales to large sample sizes and cell-type numbers

Source
scCODA is a Bayesian model for compositional single-cell data analysis.
Embedded
yes

Text

We next analyzed compositional changes of cell types in single-cell data from patients with ulcerative colitis (UC) compared to healthy donors1. Here, biopsy samples from the epithelium and the underlying lamina propria (Fig. 3c, Supplementary Data 3, and Supplementary Fig. 7) were enzymatically separated and subsequently analyzed with scRNA-seq, resulting in 51 cell types from 133 samples. The epithelium and the lamina propria represent two different compartments and were tested separately. However, some epithelial cells ended up in the lamina propria samples and vice versa. For testing, we summarized these cells as nonepithelial in the epithelium and as epithelial in the lamina propria (Fig. 3d, e). We then reanalyzed the data with the Dirichlet regression model used in Smillie et al.1, leading to more statistically significant results compared to the original publication. Similar to the Dirichlet regression model, scCODA identified several statistically credible cell-type changes in healthy tissue compared to both non-inflamed and inflamed tissue in both epithelium and lamina propria at an FDR level of 0.2, using Immature Goblet cells and CD8 + intraepithelial lymphocytes (IELs) as automatically selected