mhoban
commented
2 months ago I think the command-line options for this ought to be restructured as well. Rather than just --illumina-demultiplexed (which I always thought was a cumbersome option) or not, we should have:
--demultiplexed-by <val>
Where val can be one of
barcode
index
combined
- For the
indicesoption, assume that all fastq files already represent individual samples.- barcode file will be used for primer-mismatch check (unless not wanted)
- For the
barcodesoption, annotate samples in fastq file(s) using barcode file(s), then split annotations by sample- typically, only one (for single-end) or two (for paired-end) fastq files are given
- what happens if there are multiple fastq files?
- For the
combinedoption, primer barcode combos are reused across different index pairs- barcode file must match the first column to the prefix (before underscore) of the fastq files
- barcode file will be split into multiples, to ensure unique combos per fastq
- "pooled" fastqs will be annotated and split using the split barcode files
As written, we currently either expect combined samples multiplexed using barcoded primers OR samples separated using Illumina indices. It turns out that sometimes people do both things, pooling barcoded samples together into units delineated using Illumina indices (and reusing barcodes across index pairs). While this can get pretty sticky if folks are not extremely careful with their sample tracking, it seems to be a valid use case and the pipeline should support it.