Consequences of assembling strand-specific RNAseq as if it was not strand-specific

Question

[RNAfan]

Hi,

Recently I worked on assembling a transcriptome of a non-model organism. I was told the RNA-seq data was not strand-specific, and I assembled the transcriptome with Trinity, using the not strand-specific option.

It turns out that the reads are actually strand-specific. I would like to know what are the consequences of keeping the current transcriptome (in other words, of analyzing strand-specific data as not strand-specific).

My study organism is a vertebrate, so I don't think the genome is particularly compact. The main use of this transcriptome was to guide a genome annotation process. We are quite satisfied with the annotation derived from the current transcriptome.

I know that the best thing to do would be to redo the analyses, but this would be time consuming and would require a lot of computing resources, that we don't have right now. I guess I am looking for reasons to convince my boss that this is what we should do.

Thanks!

Answer 1

[jotan]

I doubt that strand-specificity will have a big impact on your results.

But, you absolutely have to re-do the analysis. It would be irresponsible to not correct a technical error once you know it's there.

Answer 2

[Biomonika (Noolean)]

You likely got worse assembly (than what you would get, had you specified strand-specifity). You can evaluate how much worse - for each transcript, count number of reads mapping in both directions, looking at which strand first read in a pair maps (you can find details how to do this in these posts: https://www.biostars.org/p/87097/). Since antisense transcription is very rare, this can either reassure you that your assembly is fine or give you evidence that you should redo it.