When I use q2-cutadpt demux-paired with 6 bases long barcodes, sequences are gathered in just one or two samples, and other samples hold fewer sequences.
demux-odd.qzv (298.2 KB)
I guess it's because the barcodes are too short and they appear in some sequences. So I add the first 6 bases of forward primer to every barcode, and they are combined to a 12 bases “barcode”. Then the demultiplex result seems correct.
demux.qzv (295.4 KB)
I wonder whether my guess and treat is correct. If yes, maybe I should re-process my previous short barcodes sequences.