Problem with demux

[Nicholas_Bokulich]

Hi @Linnie,

@Linnie

Given the length of my amplicon, would joining the fwd and reverse reads be of bad quality?

If it's possible to separately upload these reads into QIIME2, I would personally go that route and run through dada2 (which will join reads after QC).

It sounds like that might not be possible, though.

It is worth using your data as-is and see what happens — if the results seem rather noisy upon further inspection, it may be that the paired-end joining prior to dada2 is preventing detection of sequence errors. @benjjneb — do you have any thoughts on how read joining prior to dada2 would impact error detection?

@Linnie

It seems we cannot import the files into qiime2 with PairEndSequences because we have fwd and rev reads in each of our R1 and R2

I am not sure if this will prevent importation, but having reads in mixed orientations will cause issues for alignment and ASV/OTU detection downstream. The post that you linked to has the additional issue that reverse primers are contained in the read (which also implies that non-biological sequence is downstream of that!), which should not be an issue with your reads. So perhaps we can make something work.

Perhaps you could re-orient your reads, e.g., if the reverse primer is detected in a read then reverse complement it? (and vice versa for the reverse read). If primers are not still in reads, perhaps there is still a way (outside of QIIME2)... e.g., it sounds like you do have barcodes in-line in the reads:

1. Concatenate R1 + R2 --> "forward" reads
2. Concatenate R2 + R1 --> "reverse" reads
3. extract barcodes from "forward" reads.
4. Import into QIIME2 as paired reads.
5. demux and only read pairs in the correct forward orientation will be retained.

Would that work?

Alternatively, with a custom script you could determine whether the first X bases match an expected barcode; if yes, write out the forward/reverse reads as they are, and write out the barcode to a barcodes file; if no, determine whether the first X bases on the reverse read match an expected barcode; if yes, write out the reverse complements, and write out the barcode to the barcodes file.

I hope one of those solutions will solve your issue! Please let us know — this could help us decide how best to support sequence data following your setup in the future.