Just moved our ONT runs to duplex basecalling (Guppy 6.5.7 on a P2 Solo) and our SNP concordance vs PCR validation jumped above 99.5%, but we’re seeing more incidental human reads in metagenomic libraries. For those working near clinical space, are you updating consent language and de-ID pipelines to account for unexpected human DNA before cloud storage, or leaning on host depletion plus read filtering as sufficient?
Same here — after switching to ‘duplex basecalling’ on Guppy 6.5.7 our metagenomes picked up way more human, . We solved the consent bit with a one-line addendum and gate all uploads behind a local host scrub: minimap2 -x map-ont to GRCh38 and drop reads with MAPQ>=20, then sync only the unmapped FASTQs; it’s about 10–15 min per P2 Solo run and keeps cloud clean (GitHub - lh3/minimap2: A versatile pairwise aligner for genomic and spliced nucleotide sequences). If you need an off-the-shelf option, @marta used EPI2ME’s human depletion workflow, but I found it slower — have you tried that?
And totally understand the concern about incidental human reads. We found adding a specific clause in our consent forms helped clarify this for participants. It’s definitely a tightrope walk when balancing data richness and compliance, particularly with tools like Guppy 6.5.7.