scRNA-seq for Microcephaly Research [IV]: Dirichlet Regression for Single-Cell Population Differences.

Analysis of single-cell RNA sequencing typically includes the clustering of cells and subsequent determination of the population size of each cluster, relative to the whole. In an experimental setting, two or more conditions are compared to assess changes in cellular composition of the sampled tissue. Cluster populations are frequently normalized to the total number of cells from each replicate in order to facilitate comparisons. After normalization, they become interdependent fractions and therefore cannot be compared using individual t-tests. Here we describe the use of Dirichlet regression to compare changes in cellular composition between two or more conditions when multiple biological replicates (three or more) are sampled under each condition. We provide an example of R code to conduct a similar analysis and interpret the results.