weird kmer-content peak in RNA-seq data

[kareldegendt]

Hi all,

I have attached a pic of the kmer-content of my RNA-seq experiment.
Input was a fastq-file, 51bp reads, over 30 million reads, RNA-seq on an Illumina Hiseq.
At about the 21st position in the reads, I see the AAAAA 5-mer suddenly rising. Does anyone have a clue what might be causing that? I see it in all my samples. RNA samples were collected after arresting translation with cycloheximide.
Could it be that something is wrong with the fragment size? If so, how do I check that??

Thanks a million!

Karel

[kalidaemon]

Hi Karel
It's possible that the AAAA k-mer that you're seeing at around 21 base pairs are arrested transcripts caused by cyclohexamide treatment. You can check this possibility out by looking at the length distribution plot in FASTQC. However, if you see uniform 51 base pair reads, this doesn't mean that your AAAA k-mer is not due to terminated transcripts because you may have sequenced the 3' UTR. I would probably try to address this computationally by segregating out the reads with the AAAA k-mer at 21 bases. I would then trim them and run the shorter (ie 19 bp) and longer (ie 50 bp) reads through my analyis pipeline separately. I'm assuming the goal of your experiment is looking at genes that are transcribed rapidly or with whatever stimulus you gave your cells before Chx treatment vs slowly/not in response to the stimulus. This should tell you something about these genes.

[mparida]

Hi
I think the kmer contents of my data is pretty darn bad. Is there anyway to filter them or trim them from my data and analyze them separately.
Any useful comment is highly appreciated.