Could HUMAnN 3 be modified to handle paired-end files seamlessly like MetaPhlAn 3 already does? If each sample is about 3 GB on disk and I have about 100 samples, it’s an extra 300 GB of templorary files on disk. It could be more efficient by not requiring the files to be merged but handled inside of HUMAnN 3.
Thanks for your comment - I’ll add this as something to think about in future versions of the software. The bottleneck at the moment is that translated search tools (which HUMAnN wraps) tend not to consider paired reads since they are not aligning to genomes, which means at some point we would still need to concatenate the surviving paired reads (i.e. those not mapped upstream in the HUMAnN workflow).