Keep FastQC homopolymers or switch to fastp auto-detection?

Question

[Abieskawa]

Hi everyone, I’m improving my adapter-trimming pipeline and want to confirm if my current setup makes sense.

Current command:

cutadapt \
  -a AGATCGGAAGAG -a AAAAAAAAAAAA -a GGGGGGGGGGGG \
  -A AGATCGGAAGAG -A AAAAAAAAAAAA -A GGGGGGGGGGGG \
  -j 128 -m 5 -Q 20 -q 20 -o R1.trim.fq.gz -p R2.trim.fq.gz R1.fq.gz R2.fq.gz

I include poly-A and poly-G because FastQC reports them. My thought is: if they aren’t real adapters, paired-end alignment (5' ends were not touched) should still recover any useful sequence. I used to think fastp cannot clean all fastQC see, such as AAAAAAAAAAAA, but now I think it might be cause of different target length of A in fastp. Besides, fastp can not parallel over 16 threads.

Questions:

1. Is it safe to include AAAAAAAAAAAA / GGGGGGGGGGGG, or does that risk over-trimming real poly(A) tails?
2. should I move to fastp with automatic adapter detection? Is it necessary to change cutadapt to fastp?

Example fastp call for comparison:

fastp -i R1.fq.gz -I R2.fq.gz -o R1.trim.fq.gz -O R2.trim.fq.gz \
      --detect_adapter_for_pe 

Thanks for any advice!

Answer 1

[GenoMax]

Is it safe to include AAAAAAAAAAAA / GGGGGGGGGGGG, or does that risk over-trimming real poly(A) tails?

Homopolymers are not because of adapters. In case of poly-A they can be from mRNA tails. As for poly-G, those may be due to "no signal" in two color chemistries being equated as a G call.

If you want to keep the poly-A tails (for what ever reason) then don't trim those. Poly-G on other hand are likely the "no signal" calls that can be removed.

should I move to fastp with automatic adapter detection? Is it necessary to change cutadapt to fastp?

That is a personal preference. Either program, with right command line options, should be able to deal with adapters/homo-polymers etc.

Besides, fastp can not parallel over 16 threads.

That is not true. You need to add the following command line option

-w, --thread                       worker thread number, default is 3 

Use -w 16. That said, unless you have high performance storage available using that high a number may actually slow things down.