Should I skip trimming, and do deseq2 directly if fastqc only shows adapters.

Question

[Jiahui]

Hi, all. I am analyzing RNA-seq, and the fastqc results show everything is good, all green, but adapters exist. Do you think I can skip trim adapters and use deseq2 for my analysis?

Answer 1

[Asaf]

Read trimming is not required for mapping and quantification of RNA-seq reads at the gene level

[ATpoint]

Note that this is all based on the subread aligner. It is probably similar to something like STAR, but if you use other approaches such as salmon or kallisto then it is less clear whether this has an impact. I always clean my data of adapters if present, but mayb that's my personal paranoia. After all, it's a fast process, so what...

[dsull]

In my personal experience, pseudoalignment isn't affected by adapter contamination (so the case that an adapter causes a k-mer to be spuriously mapped to some transcript is rare [and if a few reads are wrong, so what]) but, of course, I haven't comprehensively tested all transcriptomes, all possible adapter sequences, all library preps, etc. (and the paper linked to above didn't either but it's an interesting paper that provides some useful insights).

Re: salmon, see Adapter trimming before mapping with Salmon

[ATpoint]

I was more concerned with unmapped reads due to the adapter content, but as you say, but that would only come into play if you have lots of adapters, and if that is the case the question of library integrity comes into play as in standard RNA-seq adapters should not or rarely be present.