I used --p-trunc-q 10 so that reads would be trimmed at the first fairly poor quality base, but there would still be enough errors for it to learn from. I can’t truncate the reads to the same length due to the variable size of the amplicon (~200-400 bp). I got a good amount of reads using this method, although I may try varying the q value. Do you think this is a reasonable approach. I thought I may try ITSexpress also.
Yes! That is okay — you can ignore that warning message (it is only a warning, after all — dada2 would fail if it could not learn the error rates).
Yes, that is usually what I do with ITS (use trunc-q). You will need to make sure you are getting appropriate yields after joining/merging. Check the stats file and look out for samples that lose lots of reads at the merging stage. If you are losing many reads there, it is probably a sign that you are losing longer reads,skewing your results. If you get reasonable merge yields, you should be okay to proceed. If you do not, consider adjusting trimming parameters or using only forward reads in your analysis.
Definitely give it a try... you would just use ITSxpress to trim away non-ITS sequence, and then input that to dada2.
Thanks Nicholas! Is there or will there be a minlen option on DADA2 like on the R script so that for ITS where we are trimming based on quality we can still remove spurious sequences say under 100 bp?
Cheers