Is Kallisto good for identifying non-coding RNA in a RNAseq data?

Question

[pr2009]

Hi

I generated a co-expression map using Kallisto for aligning reads. However, after completing my analysis, when I checked the gene_biotype for the gene list, I only found 3% non-coding RNA. I repeated the analysis with another data from (Skymap), I got similar results.

My question is this normal or Kallisto is not good in aligning non-coding RNA.

Thank you

[WouterDeCoster]

Kallisto uses a transcriptome genome index, so you can only detect what is present in the annotation.

[A. Domingues]

Adding to what WouterDeCoster said: (i) a lot of non-coding RNAs are relatively small, which means that depending on which sequencing protocol was they might not have been sequenced at all; (ii) they might also be repetitive leading to multimapping reads which I am not sure Kallisto would be able to handle even if they are kept by the mapper.

[Kristoffer Vitting-Seerup]

True Kallisto is better suited for long-noncoding RNAs (lncRNAs)

[ATpoint]

You typically map short RNA species sich as microRNAs against a microRNA-only reference such as miRbase to avoid axactly this probem.

[Kristoffer Vitting-Seerup]

I don't think miRNA should be a problem with standard RNA-seq as the miRNA are typically shorter than fragments analyzed due the size-selection step in the library prepreation.

[ATpoint]

Sorry, I misread the initial question, thinking that OP was quantifying miRNA-seq instead of miRNAs from standard RNA-seq. You are right, standard RNA-seq typically does size-selection for non-short RNA species.

Answer 1

[Kristoffer Vitting-Seerup]

Yes. Kallisto does not distinguish between coding and non-coding. It only requires that the transcripts are already annotated (so if you are looking for novel lncRNAs you should use a different pipeline).

What is the frequency of lncRNAs in the transcript annotation database to start with?