Hi,
Are there specific ways to analyse this kind of data ? How different is it from exon only data? Are there any specific factors to consider? Additionally, is there a way to calculate the percentage of intronic reads for each cell?If a cluster has a high percentage of intronic reads, can it be considered as an artifact?
By default in CellRanger (lets assume you have 10x data processed with it) intronic reads are included. What you get in your matrix.mtx file is the sum of all reads, including the exonic, intronic and ambiguous reads (when it doesn't know whether a read is exonic or intronic). That uses all information available for every cell.
If you want to do custom calculations you need a matrix of spliced and unspliced counts, e.g. returned by velocyto, STARsolo, alevin-fry, kallisto-bustools etc.
Whether a good proportion of unspliced counts is an artifact or not cannot be answered per se. After all, could be genuine biology, maybe it's cells actively producing RNA, that are developing towards a more mature stage which requires entire sets of genes being transcribed de novo.
Unless you want to do RNA velocity, for which it is required, I suggest you stick to standard analysis as suggested in guides such as Seurat vignettes, the Bioconductor single-cell guide OSCA, the scRNA_seq best practice book (all that can be googled) and call it a day. Tricky thing with custom analysis is that it requires a fair amount of experience to get a feeling for what is robust and meaningful rather than junk. And scRNA-seq is prone to noise, so this "junk in, junk out" thingy can easily happen.
You can calculate the percentage of intronic reads for each cell using the DropletQC package for R.
In this preprint we discuss the presence of null intronic content as well as extremely high intronic content clusters as artifacts in scRNA-seq data: https://www.biorxiv.org/content/10.1101/2024.04.18.590104v2.article-metrics
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6