This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Should I remove Kmers identified in FastQC?

Hi, apologies for this basic questions, I new in NGS quality control. I have been check my NGS data (Illumina - HiSeq 2500 2*100pb) using FastQC after trimming Nextera Adaptater with bbduck (BBTool package) using trimming overlap (ktrim=r k=25 mink=11 hdist=1 tpe tbo). And the checks fails Per base sequence content, Per sequence GC content and Kmer Content. My question is should I be trying remove the first base (~15) of my sequence? When I try this, Kmer Content fails again but at the end of the sequence. Should I remove again at the end? After my goal is to assemble them with a De Bruijn assembler and thus i'm afraid of over-represent kmer.

Thanks for the help.

enter image description here enter image description here

illumina fastqc ngs
• 5,017 views • link
updated by GenoMax GenoMax • written by Randomguy • Randomguy

2 answers

This post from Dr. Simon Andrews (author of FastQC) should be the last word about this question that keeps coming up in one form or other.

• 2 views • link
written by GenoMax GenoMax

Thanks a lot for this answer. Definitely the last word about it!

• 2 views • link
written by Randomguy • Randomguy

My plots for targeted re-sequencing with Illumina MiSeq are pretty much similar. I guess, Dr. Andrews' recommendation not to trim the first 10 something bases is applicable for my case? Сorrect me if I am wrong.

• 0 views • link
written by lamteva.vera lamteva.vera

You should you start looking at those first 10 bases only if you are having trouble with alignments, otherwise do not trim.

• 0 views • link
written by GenoMax GenoMax

The post link is not working. Could you just post the details?

• 0 views • link
written by xiaoleiusc xiaoleiusc

QCFail site is back in operation.

• 0 views • link
written by GenoMax GenoMax

In fragmentation the Nextera kit enzyme prefers to cut DNA from certain places, which is probably the reason for the bias you observe. I would let it be..

• 1 views • link
written by 5heikki 5heikki

Yup, and in NGS in general there are about a million reasons why you get over-representation of certain kmers, etc. The FastQC thresholds in my experience are really based on what you tend to see with pretty basic Whole Genome Sequencing, lots of protocols will tend to introduce biases for certain sequences appearing in reads. Unless you notice what looks like actual contamination from primers, index sequences, etc I wouldn't worry about it much.

• 1 views • link
written by DG DG

Log in to answer this question.

More posts like this