FastQC Kmer content: spike in reverse read

[reubennowell]

Hello,

I'm using FastQC to check the quality of my 250bp PE Miseq reads prior to assembly, and am seeing a spike in the relative enrichment of the 5-mer 'TTATA' around the 160-169 bp region of the reverse read (see the attached PNG). This pattern is repeated in a lot, but not all, of the genomes I'm assembling.

The reads have been put through both CutAdapt and TagDust for adapter / barcode trimming, as well as Condetri for quality trimming.

Has anyone had experience of something like this, and could enlighten me as to possible causes?

Thanks!

[Torst]

Did your adapter clipping actually include the new Nextera adaptors?

Notice Nick Loman had similar TTATA in his FASTQC analysis. Your library is too short and you are getting overlapping reads which read through into the adapter:

http://pathogenomics.bham.ac.uk/blog/2013/04/adaptor-trim-or-die-experiences-with-nextera-libraries/

[reubennowell]

Hi Torst,

Thanks for the link to Nick Loman's (excellent) blog - I think that is exactly what I'm seeing. Increasing the kmer length with the -k flag in FastQC revealed the enrichment of the sequence 'TCTCTTATACACA', the reverse complement matches exactly to the transposon sequence used in the Nextera DNA sample prep kit... (searching in the letter from Illumina given in a link in the blog above).

The centre that actually did the sequencing claimed to have removed adapter sequences (using cutAdapt) before shipping them to me, but I guess they did not include the transposon sequences in their blacklist. A friend provided me with a fasta list of a bunch of known adapters that I have fed into TagDust to do my own round of trimming, but I'm not so sure it includes the Nextera sequences...

My question now becomes: what is the best way to compile a list of sequences that can be used to search against during the trimming stage? Do you know where an up-to-date list of these sequences that includes the Nexera sequences can be found? And what is your favourite trimming tool?!

Thanks for your help

[Torst]

All the adaptors are in a PDF that you have to ask Illumina for. They recently changed their policy and allow you to dsitribute them in software.

Trimmomatic comes with them now, but you still have to put them on the command line option. Our tool Nesoni has a clip: command which has the adaptors included, and used, by default.

The MiSeq Reporter Software, if set up properly, should trim the transposon sequence. When this happens you get lots of different length reads, rather than all 150bp for example.

[reubennowell]

Cheers Torst!

Using CutAdapt with the reverse complement to the forward read transposase sequence trims a significant proportion of the R2 reads and removes the position-specific Kmer enrichment as seen in the pics I posted earlier. So I guess the insert size is small enough such that the sequencer is reading right through the insert DNA and into the stuff appended to the 5' end of the fragment. I've emailed the sequencing guys, but my guess is they forgot to include this in the trimming they did at their end (or failed to revcom it), which is why it was only observed in the R2 readset.

A quick final question: this problem became apparent when upping the default Kmer value in FastQC from 5 to 8 (as suggested in Nick Loman's blog also (he goes to 10)). But this throws up a whole bunch of apparently enriched 8-mers in the subsequent FastQC analysis after trimming the transposase sequences (see attached). There are a whole ton (~4,500) reported enriched sequences in the table outputted below the Kmer graph, and the same pattern is given for the R1 readset (i.e. it's not unique to R2).

The strange thing is that running FastQC with the default -k 5 reports no enriched kmers... now why should there be such a massive difference between the two settings? Is this a bit of a bug in FastQC?

Many thanks for your continued help!

Reuben