Abstract
Normalization with respect to sequencing depth is a crucial step in single-cell RNA sequencing preprocessing. Most methods normalize data using the whole transcriptome based on the assumption that the majority of transcriptome remains constant and are unable to detect drastic changes of the transcriptome. Here, we develop an algorithm based on a small fraction of constantly expressed genes as internal spike-ins to normalize single-cell RNA sequencing data. We demonstrate that the transcriptome of single cells may undergo drastic changes in several case study datasets and accounting for such heterogeneity by ISnorm (Internal Spike-in-like-genes normalization) improves the performance of downstream analyses.