low GC% peak in one end of paired end reads

[AlexCalderwood]

Hi,
I have paired end RNA seq data prepared from Brassica napus using TruSeq kit. After adapter trimming, FastQC shows a second low GC% peak per sequence in the _1.fq files. The _2 files all look ok.

The low GC% reads don't align to our reference transcriptome, but after blasting a small proportion of the unaligned reads, don't appear to be contamination from another organism - (hits are mostly predicted genes for Brassicas).

The average GC content is consistent across the length of the reads.

Does anyone know what might be causing this, particularly in only one of each set of read pairs?

thanks,
Alex

[pmiguel]

RNAseq?
Tell us more about the libraries.
Are you say the forward reads show this bimodal GC distribution but the reverse reads do not? Or does "_1" and "_2" mean something else.
--
Phillip

[AlexCalderwood]

Hi Phillip, thanks for your attention - what would you like to know about the libraries?

Yes exactly, the forward "_1 file" reads are red and orange lines in the thumbnail, the reverse reads are the green. Some of the reverse reads samples have a slight shoulder in the low GC region, but much more minor than the _1 files.

[pmiguel]

How were the libraries constructed? What average insert size did they have? Were the libraries stranded?

--
Phillip

[AlexCalderwood]

They were made using "NEB next ultra directional library kit", which uses dUTP method to retain strandedness, and should give an insert size of ~200bp

[pmiguel]

They were made using "NEB next ultra directional library kit", which uses dUTP method to retain strandedness, and should give an insert size of ~200bp

Okay, then my hypothesis is that the reverse read is always reading 5' in the cDNA of the forward read. So that elevated AT% is just polyA tail. Or, since you mention hits to "predicted genes", the elevated AT% may just be 3' or 5' non-translated. (Not sure which orientation the NEB kits retain.) Nor whether a 5' or 3' bias is likely in your sequence.

The non-translated regions of plants are often replete with transposable elements which can themselves have lower GC content. Or, with time after insertion, often become reduced in GC due to cytosine methylation. That is, C deamination is easily repaired because "U's" don't belong in DNA. However, 5-me-C deaminates to "T". So, over evolutionary time, simply methylating transposable elements has a sort of slow-motion "RIPping" effect.

Just speculation on my part, of course.

--
Phillip

[nucacidhunter]

Could you also post "Per base sequence content" plot form FastQC output.

[AlexCalderwood]

Please see attached for "per base sequence content" for one of the reverse read problem files post trimming. (Sorry, in a previous post I screwed up forward and reverse reads -> _1 is reverse, relative to mRNA)

I think the gradient of the GC lines is consistent with Phillip's idea of the AT rich 3'UTR being a factor.