cutadapt and paired-end reads when you don't know your adapter sequences...

Question

[beegrackle]

Hi all, I have some amplicon sequencing data from a company that has been endlessly frustrating to work with...it is a long story, but they basically dumped my sequencing data on a server without any information about what the adapter sequences were, etc. and then skedaddled.

So I went and looked at the sequences and found what I thought were the adapter sequences by hand/eyeball, and just checked that those adapter sequences I'd found were indeed in (almost) every read after the primer reverse complement at the 3' end of the purported amplified sequence. I did this separately for R1 and R2 because after the first parts of the adapter sequences (which are identical, and also match illumina adapter sequences I found online) the adapter sequences appear to diverge.

Because many of my actual amplified reads were so short, the reads had the adapter sequences and then poly-A tails and a ton of noise at the 3' end of my reads, so I didn't want to merge my paired-end reads until I'd trimmed them.

I used cutadapt to trim my sequence, and no matter what I do - use only the identical beginning part of the adapter for both, use the full 'adapters' (different for R1 and R2; I put this in quotations because it's just me eyeballing what I think it is), use 50% or 75% of the full 'adapters', even use the reverse complements of the primers - I get uneven results for R1 and R2. As in, I have a different distribution of sequence lengths for R1 and R2 after trimming. I would expect some small differences but in all of my fastq files, the R2 sequences are longer after trimming.

Has anyone dealt with this before/can suggest a better strategy? I'm thinking I might just lop off the ends of my sequences to get rid of some of the noise, merge the paired ends and then try trimming....

Answer 1

[h.mon]

You can use bbduk (from bbtools) with the flags tbo (trim adapters based on where paired reads overlap) and tpe (when kmer right-trimming, trim both reads to the minimum length of either).

[beegrackle]

Thanks - that worked out really well!

[GenoMax]

Stay with BBTools and use bbmap.sh to align your data. You will be pleased with the results :)

Answer 2

[BioinfGuru]

Don't fastqc + multiqc both return overrepresented sequences? That will tell you the adapter sequence exactly.

[beegrackle]

Unfortunately - no. Instead all my overrepresented sequences are my forward primer (or reverse primer) and a following 10-bp sequence. Which is rather dodgy, I admit.

[BioinfGuru]

I wouldnt be suprised if the company didnt provide a qc report with the data they gave you... money back...not good enough at all.