A robust statistical framework to detect multiple sources of hidden variation in single-cell transcriptomes

Single-cell RNA-Sequencing data often harbor variation from multiple correlated sources, which cannot be accurately detected by existing methods. Here we present a novel and robust statistical framework that can capture correlated sources of variation in an iterative fashion: iteratively adjusted surrogate variable analysis (IA-SVA). We demonstrate that IA-SVA accurately captures hidden variation in single cell RNA-Sequencing data arising from cell contamination, cell-cycle stage, and differences in cell types along with the marker genes associated with the source.