Unbiased quantification of spliced and unspliced reads for RNA velocity analysis

Abstract

To predict the future state of single cells, RNA velocity can be used to infer a temporal dimension in single-cell RNA data. RNA velocity uses the abundances of spliced and unspliced RNA in a cell for each gene to predict the direction of change in the transcriptional space of the cells. Inaccuracies and biases in the quantification of spliced and unspliced reads have been shown and suggested in several papers 6 & 8. For reliable velocity calculations, it is crucial that reads are quantified without bias towards either spliced or unspliced. This work introduces a counting method that quantifies spliced and unspliced reads in an unbiased way. The method is tested on a 10x dataset, and on a full-length sequencing technique called Vasa sequencing. The difficulties of using RNA velocities with the current sequencing technologies are shown and the importance of filtering transcript models before the analysis is highlighted. Lastly, some potential future steps are suggested.