paperKB
coga / coga-kb
Help
Sign in

Chunk #58 — Methods — Analysis of publicly available datasets — Single-cell RNA-seq data of ulcerative colitis in human donors

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

Text

The original study annotated all cell types together, resulting in 51 different cell types. However, some cell types that are originally located in the LP have been found in the epithelial samples and vice versa. For the differential composition analysis of the Epi and LP, we considered the nonepithelial and epithelial cell types, respectively, as one group. Therefore, we tested the changes in 16 cell types in the Epi and 37 cell types in LP. In addition, we reanalyzed the data using the Dirichlet regression model as in Smillie et al.1 (with R package DirichletReg v.0.7-0 in R v.3.5.2). Importantly, we realized that Smillie et al. summed up the counts of the same replicates (as described in the analysis scripts in https://github.com/cssmillie/ulcerative_colitis), while we consider every replicate as an independent. Overall, we have data from 29 donors (61 samples, where 24 healthy, 21 non-inflamed, 16 inflamed) in Epi and data from 30 donors (72 samples, where 24 each healthy, non-inflamed, and inflamed, respectively) in LP. Specifically, we tested chain lengths of 20,000, 40,000, 80,000, and 150,000 iterations with a burn-in