Interpreting 5'/3' bias

[sidderb]

Hi All,
We have 100bp PE RNAseq data generated on a HiSeq2000 from a library prepped with the Nugen Ovation v2 kit. The samples are from human and rat, with the human samples of lesser quality (RIN<7) whilst the rat are fine. Reads were aligned with bowtie using default options.

Using RSeQC we see the following profiles of counts across transcripts:



Does anyone have any ideas on how to interpret this? I was expecting a possible 3' bias from the degraded(?) samples and/or the library prep, but wasn't expecting to see both a 3' and 5' bias...

What would you do on the back of this, anything?

Thanks.

[sidderb]

Can anyone help?

[aparna]

What %of reads were mapped?

[sidderb]

~60% in all samples

[scooter]

I suspect the library prep method was not a stranded protocol. This would result in ~1/2 of the PE data being in one orientation (i.e. reads are 5'-3' relative to transcript) and the other 1/2 being in the opposite orientation (3'-5'). I have heard the ovation kit has a 3' bias, so in your case the 5' bias is simply the reads in the antisense direction.

I am not sure this is correct, just a hunch.

[aparna]

I would suspect at the library prep phase as scooter.But since the data is out - I would try to improve mapping by using other aligners.
Bowtie is good for RNA seq with end to end mapping (sensitive mode) but in that way it is more stringent and I personally dont want to use it as my first aligner of choice.
If you like Bowtie's performance - I would suggest you to use Bowtie2 - which allows gapped alignment.
I would also suggest you to use bwa just to have another comparison of mapping.

By the way have you done pre-mapping QC?Did you had to trim the reads?
what were you mapping to?

-Aparna

[pmiguel]

Hi All,
We have 100bp PE RNAseq data generated on a HiSeq2000 from a library prepped with the Nugen Ovation v2 kit. The samples are from human and rat, with the human samples of lesser quality (RIN<7) whilst the rat are fine. Reads were aligned with bowtie using default options.

Using RSeQC we see the following profiles of counts across transcripts:

[ATTACH]1333[/ATTACH]

Does anyone have any ideas on how to interpret this? I was expecting a possible 3' bias from the degraded(?) samples and/or the library prep, but wasn't expecting to see both a 3' and 5' bias...

What would you do on the back of this, anything?

Thanks.

Doesn't this kit do ligations to the end (or extensions from the end)? If so, it might be the result of a failure of the RNA in the human sample to fragment prior to adding adapters.

If you had provided a link to the protocol I might have read it and been able to give you a more targeted answer.

--
Phillip

[liguow]

Hi All,
We have 100bp PE RNAseq data generated on a HiSeq2000 from a library prepped with the Nugen Ovation v2 kit. The samples are from human and rat, with the human samples of lesser quality (RIN<7) whilst the rat are fine. Reads were aligned with bowtie using default options.

Using RSeQC we see the following profiles of counts across transcripts:

[ATTACH]1333[/ATTACH]

Does anyone have any ideas on how to interpret this? I was expecting a possible 3' bias from the degraded(?) samples and/or the library prep, but wasn't expecting to see both a 3' and 5' bias...

What would you do on the back of this, anything?

Thanks.

Maybe you can try "read_distribution.py" in RSeQC package to reaffirm this observation. You are expecting to see higher coverage in both 3'UTR and 5'UTR than the internal CDS exons.

[pbluescript]

Did you map to a genome or transcriptome?
If you mapped to a genome, you should try mapping with something specifically made for RNA-Seq data like Tophat or STAR so that you can map reads split across an exon-exon junction. I have noticed a larger 3'UTR bias sometimes when I map RNA-Seq data with Bowtie or BWA to a genome. The 3'UTR is often the longest part of a gene and you'll get more reads mapping there since there are fewer split reads.

[kopi-o]

I have seen a similar large 3' bias in rRNA-depleted libraries, but not poly-A selected ones. I don't know how to explain it, though.