iDESC: identifying differential expression in single-cell RNA sequencing data with multiple subjects

BMC Bioinformatics

Table 1 Overview of the twelve DE analysis methods for comparison

	Dropout	Subject effect	Test	Model
iDESC	✓	Mixed model	Wald test	Zero-inflated negative binomial mixed model
MAST-RE	✓	Mixed model	Likelihood ratio test	Two-part hurdle mixed model
Muscat-MM	×	Mixed model	Wald test	Negative binomial mixed model
Muscat-PB	×	Aggregation	Quasi-likelihood F-test	EdgeR on sample-level aggregated data
subT	×	Aggregation	Student’s T test	T test on sample-level aggregated data
DEsingle	✓	×	Likelihood ratio test	Group-specific zero-inflated negative binomial model
MAST	✓	×	Likelihood ratio test	Two-part hurdle model
scDD	✓	×	Kolmogorov–Smirnov test	Dirichlet process mixture of normals
NBID	×	×	Likelihood ratio test	Negative binomial model with group-specific dispersion
DESeq2	×	×	Wald test	Negative binomial model with the same dispersion between groups
limma	×	×	Moderated T test	Linear regression model
Wilcoxon	×	×	Wilcoxon rank sum test	Nonparametric test

ISSN: 1471-2105