remove over-represented sequences

Question

[wuzongze001]

I download data SRR1029258.sra (GSM1263454) from GEO and run fastq-dump to convert it to fastq. Then I use fastqc to do quality control. I found the quality of sequences is not high enough and the highest over-represented sequence percentage is about 29.9%. So I use command:

fastq_quality_trimmer -t 10 -l 40 -i SRR1029258.fastq -o trim_SRR1029258.fastq

The per base sequences quality is better but the highest over-represented sequence percentage turn to be 33.1%.

According to google, over-represented sequences may due to contaminated and may cause a wrong conclusion. Is there any way I can remove these reads?

Answer 1

[Devon Ryan]

Given that this is a ChIPseq experiment, over-represented sequences are expected. If you were to completely remove these, then you'd end up removing at least some of the real peaks.

[wuzongze001]

Thank you very much! I suddenly realize that.

Answer 2

[Adrian Pelin]

rRNA is a common overrepresented sequence, in some of the RNA-Seq reads i'm working with I have contamination ranging from 15% to 22%. The easiest thing to do is to blast your overrepresented sequence. Does FastQC say No Hit next to it?

[wuzongze001]

I blast it, top two hit both are rRNA. And the top five over-represented sequences percentage of my fastqc result are 33%, 4.5%, 3.6%,1.9%,0.6. I do not really understand "Does FastQC say No Hit next to it" , could you explain this?

[Devon Ryan]

FastQC has a limited database of things that it checks against, it won't be as thorough as blast.

[Adrian Pelin]

Devon has a point, having the entire rRNA database in the package would make the software large in size and difficult to download. Although I must say, I haven't figured out how to easily cut and paste the sequences that are overrepresented, I usually have to save the report and then open the HTML, to copy the sequence.

[lawrence.mckechnie]

Hi! I run fastqc and I got a bunch of over-represented sequences with "no hit" next to them... I'm not sure if these are adaptors or important biological information that should stay in there? I guess what I'm really asking is the following: what is the meaning of "no hit" in this context? I'm going to try removing them (using cutadapt) and see if the per-base quality improves (I was getting really low quality at the ends of reads). Any advice anyone could give me regarding this issue would be really appreciated. Cheers!

[Adrian Pelin]

"The easiest thing to do is to blast your overrepresented sequence"

Answer 3

[Gary]

If the percentage of unique reads is too low, it could be caused by PCR bias during ChIP-Seq library preparation.