I have recently received amplicon sequencing data generated on the Illumina MiSeq i100 platform (16S rRNA gene, 515F/806R primer pair). When importing and visualising the demultiplexed reads in QIIME 2, the interactive quality plots show high and constant quality scores at Q38.
Given this unusual lack of quality variation and decay, I am unsure how appropriate it is to proceed with the standard DADA2 denoising workflow in QIIME 2.
In particular, I have two main questions:
Is it valid to proceed with DADA2 denoising on MiSeq i100 data in the usual way?
Without the typical quality score drop-off, what is the recommended approach for determining truncation parameters?
Any guidance or shared experience would be greatly appreciated.
Thank you very much for your answer and for pointing me towards the post addressing binned quality scores. That all helped a lot to understand that it is fine to proceed with the usual denoising steps.
After primer removal, my sequences were roughly 273nt long and in I ran DADA2 with different truncation lengths. No truncation - or selecting 270/270 - yielded very poor scores (~17.5% of input non chimeric), while truncating at 250/250 - or shorter - had much better outcomes (~86.8% of input non-chimeric). It thus seems that the final bases may not be of great quality after all, or that they don't belong to the biological insert, which would create weird sequences identified as chimeras.
It thus seems that the final bases may not be of great quality after all, or that they don't belong to the biological insert, which would create weird sequences identified as chimeras.
This could be due to poly-G tails which are characteristic of sequencing runs on newer illumina machines like the MiSeq i100. You can use the --p-nextseq-trim parameter in qiime cutadapt trim-paired to deal with these more precisely than using truncation if you'd like.
