A framework for designing splice-junction experiments in deep 3' single-cell RNA sequencing

Alternative splicing, a process specific to certain cell types and relevant to disease, has not been optimized for in single-cell RNA sequencing, which is primarily designed for gene-level quantification. The study aimed to quantify how sequencing depth affects the recovery of splice-junctions in the dominant 3 -biased chemistries.

To achieve this, researchers generated deep single-cell RNA sequencing data from human airway epithelial cells, resulting in 12 samples with read counts ranging from 102 to 803 million per sample. This data was then compared to matched bulk RNA sequencing data from a cohort of 190 donors. The findings showed that junction detection did not reach saturation with increased depth; instead, read support rather than annotation status determined the validity of the results.

To recover 90% of the observed full-depth junction complement, approximately 52,250 reads per cell were required, which is significantly higher than what is needed for gene-level analysis. Quantifying splicing proved to be more depth-limited compared to detecting it, with rare cell types being limited by the number of cells rather than sequencing depth.

When compared to matched bulk RNA sequencing, single-cell recall plateaued at a practical ceiling determined by sample size and sparse sampling, and was found to be independent of the distance from the 3 -end. These results not only provide a framework for designing splicing experiments in deep 3 -biased single-cell RNA sequencing but also offer an airway-epithelial dataset paired with matched bulk data for benchmarking splicing methods.