How much adapter contamination is common?

[knostrov]

Hi everybody,

I am looking at my Fastq QC stats with much surprise: 25% of my reads contain Trueseq adapter sequences. That seems like a lot. I am completely new to NGS data analysis, so I have no point of comparison... is this very unusual? What type of analysis should I do to find out more about the roots of this problem?

This is a 50bp PE run on a HiSeq 2000, target enrichment using Haloplex.

Any comments would be appreciated.

Thanks!

[kopi-o]

It sounds like a lot. Is it a special protocol, like small RNA sequencing? With very short insert sizes, you expect to see a lot of adapters in the reads. Maybe you could start by obtaining fragment size statistics from the lab.

[Wallysb01]

That is a lot. In my experience it is <1%.

[jimmybee]

Its very protocol specific. Some can have massive amounts of primer left after sequencing, so maybe fill us in on the sequencing experiment and what you did before sequencing?

[knostrov]

Thank you for the quick replies. Actually I have by now realized that almost all of these contaminating sequences are reads that consist of nothing but adapter sequence, e.g. 50bp of read2-primersite--tag--adapter. I imagine those must be two linked adapters with no insert in between. One would think such constructs would be eliminated by size selection? Also their abundance varies per sample. They make up between 2-25% of all reads, depending on the sample.

So the specifics of this sequencing experiment are (I did not do the experimental part myself...)
- FFPE tissue samples, good DNA quality according to size distribution on gel
- target enrichment using the Haloplex protocol (custom probes that hybridize to the two ends of a desired region to make a circle --> ligation --> circle amplification)
- 50 bp PE run on HiSeq 2000

[MWN]

Thank you for the quick replies. Actually I have by now realized that almost all of these contaminating sequences are reads that consist of nothing but adapter sequence, e.g. 50bp of read2-primersite--tag--adapter. I imagine those must be two linked adapters with no insert in between. One would think such constructs would be eliminated by size selection? Also their abundance varies per sample. They make up between 2-25% of all reads, depending on the sample.

So the specifics of this sequencing experiment are (I did not do the experimental part myself...)
- FFPE tissue samples, good DNA quality according to size distribution on gel
- target enrichment using the Haloplex protocol (custom probes that hybridize to the two ends of a desired region to make a circle --> ligation --> circle amplification)
- 50 bp PE run on HiSeq 2000

Any update?
How much FFPE DNA do you use? Appreciate.

[swNGS]

For Haloplex, you really need to do adapter trimming before alignment, then clip 5 bases of the ends of each read, otherwise it suffers from biasing of the wild type allele if you get a variant at a restriction siteor probe binding site.
Yes, you should be able to reduce them by size selection, but as a previous person said, have a look at the range of insert sizes (after establishing that trimming was done correctly, otherwise you'll lose a lot of potentially useful sequence). We are using Halo with 150 bp reads and I suspect we've got a bigger adapter read-through problem than you!