Strange looking fastQC per base sequence content, duplications, overrepresented, KMER

[aurorasea1]

Hello, I am new to analysing RNAseq data and I'm learning how to run bioinformatics just by self-reading.
The samples were paired-end reads, sequenced by Illumina HISEQ 2000 and then I ran the FastQ files via FastQC and I generated the following report:

Sequence tenth: 101
I received a "RED CROSS" for per base sequence content and I've attached the graph for your information.

I am so received a "RED CROSS" for sequence duplication levels.

The over-represented sequences are:
GTGGGTGGTGGTGCATGGCCGTTCTTAGTTGGTGGAGCGATTTGTCTGGT
GGGTGGTGGTGCATGGCCGTTCTTAGTTGGTGGAGCGATTTGTCTGGTTA
GCGAATGGCTCATTAAATCAGTTATGGTTCCTTTGGTCGCTCGCTCCTCT
However, I have no idea where they're coming from because they do not match my adapter sequences (GGAGAA) and also FasQC does not report them as 'adapter sequences'.

I'll show the K'mer content graph as well, with most over-represented kMERS at the 3' end.

I wonder what could be the source of these over-represented sequences and how best to deal with them?
Thank you very much for your advice.

[nucacidhunter]

They are 100% match to the following:
GTGGGTGGTGGTGCATGGCCGTTCTTAGTTGGTGGAGCGATTTGTCTGGT
PREDICTED: Sinocyclocheilus rhinocerous neuroligin-3-like (LOC107733424), transcript variant X14, misc_RNA

GGGTGGTGGTGCATGGCCGTTCTTAGTTGGTGGAGCGATTTGTCTGGTTA
PREDICTED: Capsicum annuum uncharacterized LOC107852466 (LOC107852466), ncRNA

GCGAATGGCTCATTAAATCAGTTATGGTTCCTTTGGTCGCTCGCTCCTCT
Uncultured eukaryote 18S ribosomal RNA, partial sequence

Red crosses are intended to indicate shotgun library quality and most of the time they will show up in RNA-Seq data. Duplication rate is high but that would be dependent on the quantity of input RNA.

[aurorasea1]

They are 100% match to the following:
GTGGGTGGTGGTGCATGGCCGTTCTTAGTTGGTGGAGCGATTTGTCTGGT
PREDICTED: Sinocyclocheilus rhinocerous neuroligin-3-like (LOC107733424), transcript variant X14, misc_RNA

GGGTGGTGGTGCATGGCCGTTCTTAGTTGGTGGAGCGATTTGTCTGGTTA
PREDICTED: Capsicum annuum uncharacterized LOC107852466 (LOC107852466), ncRNA

GCGAATGGCTCATTAAATCAGTTATGGTTCCTTTGGTCGCTCGCTCCTCT
Uncultured eukaryote 18S ribosomal RNA, partial sequence

Red crosses are intended to indicate shotgun library quality and most of the time they will show up in RNA-Seq data. Duplication rate is high but that would be dependent on the quantity of input RNA.

Thanks very much but I'm very surprised how the first 2 ended up in my sequence (!)
But would they cause problem in my subsequent analysis? Because I'll be mapping them to the human genome rather than de novo assembly, I suppose that'd exclude these 'non-mappable' reads?
Thank you

[Michael.Ante]

Hi,
it looks, as you have a random-primer in the first 8-9 bases. It might be useful to either trim this sequence or increase allowed mismatches in your alignment.

Regarding the over-represented sequences, they might or might not map to your target genome. Nevertheless, if you have low alignment rates, you may want to check the unmapped reads. Some aligners like TopHat2 store these automatically, others like the STAR aligner need a parameter for saving them.

Cheers,

Michae

[GenoMax]

Best solution may be to do nothing (except scan for presence of adapters and remove them if necessary).

Take a look at these two posts. RNAseq bias and Over-representation.