ZIAQ: A quantile regression method for differential expression analysis of single-cell RNA-seq data

MOTIVATION: Single-cell RNA sequencing (scRNA-seq) has enabled the simultaneous transcriptomic profiling of individual cells under different biological conditions. scRNA-seq data have two unique challenges that can affect the sensitivity and specificity of single-cell differential expression analysis: a large proportion of expressed genes with zero or low read counts ("dropout" events) and multimodal data distributions. RESULTS: We have developed a zero-inflation-adjusted quantile (ZIAQ) method, which is the first method to account for both dropout rates and complex scRNA-seq data distributions in the same model. ZIAQ demonstrates superior performance over several existing methods on simulated scRNA-seq datasets by finding more differentially expressed genes. When ZIAQ was applied to the comparison of neoplastic and non-neoplastic cells from a human glioblastoma dataset, the ranking of biologically relevant genes and pathways showed clear improvement over existing methods. AVAILABILITY: ZIAQ is implemented in the R language and available at https://github.com/gefeizhang/ZIAQ.