How to rerun Guppy Basecalling (with both fast5_pass and fast5_fail)

Question

[Ludwig Kian Soon]

Hi, I recently started a nanopore sequencing and used the live-basecalling feature in mk1c. It was proceeding very slowly. I want to free up the machine to do more sequencing and am thinking of recovering the raw data in queued reads and basecall it elsewhere, but find out that only a newer version of guppy is available there, and when I tested just a single passed fast5, the output generated is different (even though the model used, the min_qscore and other parameters as much as I am aware of are the same), so my question is:

1. How I can rerun the basecalling using the newer version of Guppy with the fast5 that were been split into pass and fail?
2. The GPU has got some memory constraint, I tested that I can still do basecalling without demultiplexing at one go, should I do demultiplexing after basecalling with fast5 recovered from raw or the other way around?
3. Is there a way to combine the fast5 split into pass and fail into those "raw" fast5?

[samuel.a.odonnell]

For the question about rebasecalling both pass and failed fast5; isn't it just a means of setting the input directory as the place with the both pass and fail fast5 directories and making it look recursively for fast5 data (--recursive)?

I am not sure I understand your question about demultiplexing.

For combining the fail and pass fast5, it can just be combined into a single directory, the only difference between the fast5 files is that the resulting fastq score for the reads were below an average quality threshold. Not sure I understand the question.

[Ludwig Kian Soon]

Hi samuel.a.odonnell,

Ah, I see; I didn't know it'd be that easy. When I tested the sequences in the re-basecalled fastq file is slightly shorter than the sequences in originally base-called fastq (even though I am using the same model), I thought the fast5 moved into the pass/fail folder underwent some sort of processing. Reading back at the manual, I see now that's not the case.

Regarding the demultiplexing part, I am doing sequencing 24 samples with the barcoding kit, and there's an option in guppy_basecaller to automatically split the basecalled data to different folder according to the barcodes. Just looked at the command line parameters, I think it comes after the guppy_basecaller.

Thanks for your help!