AbstractA primary challenge in the analysis of RNA-seq data is to identify differentially expressed genes or transcripts while controlling for technical biases present in the observations. Ideally, a statistical testing procedure should incorporate information about the inherent uncertainty of the abundance estimates, whether at the gene or transcript level, that arise from quantification of abundance. Most popular methods for RNA-seq differential expression analysis fit a parametric model to the counts or scaled counts for each gene or transcript, and a subset of methods can incorporate information about the uncertainty of the counts. Previous work has shown that nonparametric models for RNA-seq differential expression may in some cases have better control of the false discovery rate, and adapt well to new data types without requiring reformulation of a parametric model. Existing nonparametric models do not take into account the inferential uncertainty of the observations, leading to an inflated false discovery rate, in particular at the transcript level. Here we propose a nonparametric model for differential expression analysis using inferential replicate counts, extending the existing SAMseq method to account for inferential uncertainty, batch effects, and sample pairing. We compare our method, “SAMseq With Inferential Samples Helps”, or Swish, with popular differential expression analysis methods. Swish has improved control of the false discovery rate, in particular for transcripts with high inferential uncertainty. We apply Swish to a singlecell RNA-seq dataset, assessing sensitivity to recover DE genes between sub-populations of cells, and compare its performance to the Wilcoxon rank sum test.