Benchmarking integration of single-cell differential expression

Abstract Integration of single-cell RNA sequencing (scRNA-seq) data between different samples has been a major challenge for analyzing cell populations. However, strategies to integrate differential expression (DE) analysis of scRNA-seq data remain underinvestigated. Here, we benchmarked 41 methods for integrative DE analysis of scRNA-seq data. Batch-effects, sparsity of data, and heterogeneity of samples substantially impacted the performance of DE analysis. Several methods that yielded high performances were suggested based on various simulations and real data analyses. In particular, the bulk RNA-seq tool edgeR incorporating the observation weights and the scRNA-seq tool MAST showed overall good performances. Remarkably, analysis for a specific cell type outperformed that of large-scale bulk sample data in prioritizing disease-related genes and pathways.