Unspliced matrix from RNA velocity is it similar to pre-mRNA alignment?

Question

[akilabioinfo]

Hello All,

I have generated spliced and unspliced matrices using kallisto bustools (lamanno). My question is that the unspliced matrix is similar to the count matrix obtained from pre-mRNA and the spliced count matrix is similar to mRNA?

If so, Then can I use the unspliced count matrix for DEG analysis

Please help me !!!

Thanks

Answer 1

[dsull]

Yes, you are correct.

However, I'm not sure why you want to use the unspliced counts for differential expression analysis. What will such an analysis tell you? If a gene's pre-mRNA is higher in certain cells, it could be because of many things: e.g. because that gene is more highly expressed, those pre-mRNAs have not been spliced yet, etc..

I recommend just running an RNA velocity workflow.

[akilabioinfo]

Thanks for your reply. Maybe my question is very naive. Previously i quantified using cell ranger pipeline I am new to kallistobustools. I have quantified the counts (Kallisto bustools) spliced, unspliced region. Now, i would like to preprocess the counts and do typical integration, and clustering followed by DEG analysis and RNA velocity.

1. Now which count matrix should i use for DEG analysis? (I would like to perform DEG using both mRNA(EXON) and pre-mRNA).

adata.h5ad --> counts present in spliced and unspliced region

layers spliced--> sparse matrix spliced count matrix

layers unspliced --> sparse matrix unspliced count matrix.

1. Do we need to do doublet detection after clustering?? (removing low quality cells)

Please help me !!!

Thanks Akila

[dsull]

1. For differential expression, use both the spliced+unspliced counts (in fact, you can sum up the two to obtain "obtain total nuclear transcripts")
2. Yes, you probably should do post-alignment QC, especially if you see weird stuff (e.g. weird clustering, some cells with an abnormally high number of genes detected, etc.)