Is it normal for RCorrector to remove millions of reads?

Question

[nina.maryn]

I'm trying to build De Novo transcriptomes for unsequenced plants to do sequence analysis.

I'm trying to choose a tool for my first pass of quality filtering after running FastQC on my raw reads. I've tried AfterQC and RCorrector. It looks like AfterQC sort of made my read quality worse, at least at the ends. And RCorrector looks great, but I went from 30 mil reads to 3.5 mil reads? There were far fewer reads tossed out with AfterQC, but my understanding is that that tool is better for genome sequencing. Any advice or thoughts?

[Dunois]

What did your initial FastQC plots look like? And did you pool the reads you fed Rcorrector by any chance?

Answer 1

[ponganta]

No, this is not normal, as rCorrector should not remove any reads (if you didn't turn on -trim!). All it does is correction, and flagging of uncorrectable reads. Do you have paired end (PE) or single end (SE) reads? Did you conduct any sort of filtering prior to error correction? I'd take a look at the approach taken by the Oyster River Protocol, I found this to work quite well.

For trimming, I would rely on TrimGalore rather than rCorrector.