Nanopore 16s long reads

[jemimahair]

My name is Jemima and I recently started a PhD working on stool and vaginal samples from a clinical trial in SSA. I am currently working on sequencing the full 16s gene on nanopore platform. I have seen in the chats that other people have asked whether the mothur pipeline can be adapted for this purpose. It has been a few years since these comments and I wondered if you had any updates/ whether you think its possible since the error rate has improved significantly?

Also, if you still would like mock community data, we only run the Zymo, but happy to share if this is of interest to you?

[krmaas]

I'm working on this at UConn MARS. Once I have a functioning process in mothur, I'll share it.

[AlexandreThibodeauUdM]

Hello. old post but I will still write it. Tried almost everything in Mothur. Things is with full 16S Nanopore reads, you do not get a lot of unique, hell even preclustering to 1% of read lenght hardly do something. I do not think Mothur is made to handle so many long sequences, even though I would love to incorporate some of Mothur protocol into full 16S sequencing such as alignment.

Switch to a different tool for full nanopore 16S, get very good mock comminities, but when I look at my samples, still get some weirds vibes from some species classification.

Best of sucess

[krmaas]

Hello. old post but I will still write it. Tried almost everything in Mothur. Things is with full 16S Nanopore reads, you do not get a lot of unique, hell even preclustering to 1% of read lenght hardly do something. I do not think Mothur is made to handle so many long sequences, even though I would love to incorporate some of Mothur protocol into full 16S sequencing such as alignment.

Switch to a different tool for full nanopore 16S, get very good mock comminities, but when I look at my samples, still get some weirds vibes from some species classification.

Best of sucess

What are you using? The error rate for nanopore is higher than their q-scores suggest. When I look at just zymo mock communities with reads that are >q30 (super high accuracy bc), I'm getting ~3% actual errors. I tried generating some UMI reads (following a paper from Zeills lab at UBC). But couldn't get their bioinformatic tools to work so trying to build something on my own which is slow.

I'm having decent results for CO1 for insects processed in dada2 but species are >6% different so more forgiving for aggressive preclustering.