Hi. I'm processing a set of 80 16S sequencing samples using the ampliseq pipeline. I've generated data using the default parameters. DADA2 only trimmed off 1nt from R1 3' and none from R2. I guess the recovery rates are not that bad, but I'd like to try trimming more since fastqc shows big drops of quality scores in the 3' ends. I wonder if anyone can provide some guidance or share experience about how to determine the length of trimming. Is there a typical trimming length? Thanks.

Besides the method implemented in the pipeline, I think eye balling graphs like multiqc/fastqc is the way to go. Also, consider how long you expect the amplicon to be so that you don't start losing the longer ones in the merging step by cutting the reads too short.

Thanks for the response! I think I'll try some different lengths to see how the results look.

What I did -- a long time ago -- was look at the length distributions of the final denoised and merged sequences and how they were affected -- particularly at the longer side -- by different trimming lengths. When I saw that the distribution shortened I realized I was cutting the reads too short.

I typically rather look at the DADA2 quality plots for finding manual trimming params if needed. Usually it isnt though.

Thanks for all the recommendations!