Liang Chen, Ph.D.

Associate Professor, Department of Biological Sciences, University of Southern California

“Tackling overdispersion in RNA-seq data analysis”

The rapid advances in high-throughput sequencing technologies provide us an opportunity to dissect transcriptomes with unprecedented resolution. However transcriptome quantification is still hindered by non-uninform read sampling. Existing methods assume a constant bias factor for each relative position of genes or simply correct the sequence-specific bias caused by random hexamer priming. However, the overall bias is complicated and caused by multiple factors including many unknown ones, and the bias pattern can vary significantly across different regions and different protocols. In light of these facts, we proposed to use the generalized-Poisson (GP) model to estimate the bias in a data-adaptive way without any presumption. We further incorporated this data-adaptive bias correction in the deconvolution of isoform expression. Our methods significantly improve the quantification of isoform and gene expression as well as the derived exon inclusion rates. For single-cell RNA-seq data, our method distinguishes bias heterogeneity from true biological heterogeneity and uncovers smaller cell-to-cell expression variability.