Barcoding and training

[tnn111]

I would like to build some custom models using bonito. As input, I would like to use reads from samples that were barcoded. Is there a standard way of doing this?

Based on information that I could easily find, a way of doing this appears to do demultiplexing using guppy, collect the read IDs corresponding to each barcode and extracting the fast5 reads using the IDs and feeding those into bonito.

Is that how everyone does it or is there a better way? Is there anything written on this?

Thanks, Torben

[aistBMRG]

I guess you could use porechop or qcat for demultiplexing after basecalling. Also, guppy_barcoder is possible but needs the headers of the fastq/fasta from bonito to be updated with runid= information, as far as I understand as dummy name will work.

Any other thoughts from others, I am exploring the same issue. I tried the above approaches but all yield quite different number of reads after demultiplexing.

Regards,

Dieter

[CWYuan08]

Hi @aistBMRG, I am wondering how to update the headers of the bonito fastq with runID=information, could you please share your method? Thank you very much!