non-exonic reads in RNAseq

Question

[kanwarjag]

I understand in RNAseq we will always get some level of reads mapping to areas . However I am seeing almost 85% of reads align to outside coding region. i was wondering what may be the cause of this. Does that mean a contamination has over- powered actual data.

[Benn]

Are you using the correct reference? ;-)

It sounds like a wet lab problem, are you sure you have removed DNA from your samples?

[kanwarjag]

yes correct genome for aligning. I am looking and the distribution of reads in Bam file

[Devon Ryan]

Almost every time this happens it's because someone's using the wrong genome in IGV :)

[Martombo]

you define as "intergenic" everything that couldn't be assigned to a gene? what about intronic regions? this could eventually be also caused by looking at the wrong strand, or using a different genome version for the annotation and the mapping. did you already check these cases?

[WouterDeCoster]

Is your genome of interest/specie well characterized?

[kanwarjag]

mm9 that is well characterized

[Devon Ryan]

Are these ribo-depleted samples or polyA-enriched? If the former, then perhaps you're seeing expressed repeat regions (there can be quite a few).

[kanwarjag]

ribo depleted. I think it is sample preparation but I want to make sure

[Devon Ryan]

Try aligning against the Rn45S sequence and see how many reads you get. I routinely do that with our rRNA depleted datasets to see how depleted they actually are. My guess is that Michael is right you're getting a bunch of rRNA (and probably tRNA). Just look at some of the higher-coverage areas on the UCSC genome browser with the repeatmasker track enabled. I suspect that'll be illuminating.

[Devon Ryan]

Does anyone know why this was deleted?

[WouterDeCoster]

Nope, can we restore it?

[GouthamAtla]

I opened it but I guess the OP deleted this,

Answer 1

[Michael]

Hi Kanwarjag,

However I am seeing almost 85% of reads align to intergenic region.

This is imo extremely unlikely and I have never seen that, even with our imperfect annotation of our model the salmon louse. I have done a little randomization experiment on our data by placing random gene models in intergenic regions, and found that for most samples the 99% confidence level background read-count to experience in intergenic regions is 1. I don't have exact figures for reads overlapping intergenic regions though, but 85% is very high. I would check the following

• correct annotation version
• missing ribosomal genes from the annotation and high level of rRNA
• draft annotation with a large number of truncated gene models and missing or truncated UTR's (add some kb flanks to genes and check again)
• leak DNA

Answer 2

[Michele Busby]

Try running it through RNA Seq QC. It take a while to set up but once you do you have it. It will tell you if it's rRNA or intergenic vs intronic.