Confronting false discoveries in single-cell differential expression

Abstract

Differential expression analysis in single-cell transcriptomics enables the dissection of cell-type-specific responses to perturbations such as disease, trauma, or experimental manipulation. While many statistical methods are available to identify differentially expressed genes, the principles that distinguish these methods and their performance remain unclear. Here, we show that the relative performance of these methods is contingent on their ability to account for variation between biological replicates. Methods that ignore this inevitable variation are biased and prone to false discoveries. Indeed, the most widely used methods can discover hundreds of differentially expressed genes in the absence of biological differences. Our results suggest an urgent need for a paradigm shift in the methods used to perform differential expression analysis in single-cell data.

Diversity of Reactive Astrogliosis in CNS Pathology: Heterogeneity or Plasticity?

“…To meet these needs will require a combination of (1) increasing throughput of single-cell biology assays to achieve robust results across experimental conditions and biological replicates ( Cao et al, 2020 ), (2) the application of multi-omic approaches to the context of CNS pathologies to avoid undue reliance on the functional interpretation of single-cell transcriptomes ( Cao et al, 2018 ), and (3) bioinformatic methods development to aid biological investigators in differentiating cell type from cell state ( Butler et al, 2018 ), to prioritize cell subsets most involved in the disease of interest ( Skinnider et al, 2021b ), and to avoid false positives that have the potential to drive research in unfruitful directions ( Squair et al, 2021 ). Finally, these data will need to be met with stringent reporting standards to enable cross-disease investigations and to understand the conserved and differential responses of astrocytes across neurological disease.…”

scPower accelerates and optimizes the design of multi-sample single cell transcriptomic studies

“…Two benchmarking studies, one applying the aforementioned simulation tool and one using different example data sets, demonstrated that the "pseudobulk" approach in combination with classical differential gene expression methods such as edgeR 44 and limma-voom 45 outperforms single cell specific methods and mixed models in multi sample DE analysis 43,46 . The pseudobulk approach approximates cell type specific gene expression levels for each individual as the sum of UMI counts over all cells of the cell type and was also successfully applied in different single cell eQTL studies [47][48][49] .…”

Single cell profiling of Hofbauer cells and fetal brain microglia reveals shared programs and functions

“…Differential gene expression analysis: Differentially expressed genes (DEGs) between males and females were calculated by aggregating counts by cluster and by individual into a pseudobulk count matrix (as implemented by the Libra R package (Squair et al, 2021)), and using the Wald test for differential expression in DESeq2 (Love et al, 2014). Recent benchmarking studies have demonstrated that bulk RNAseq methods perform equally well to those developed for scRNAseq while also controlling false discovery rates (Soneson and Robinson, 2018;Squair et al, 2021). For our hypothesis-generating analyses (e.g.…”