Dorado basecalling: handling fail, pass, and skip reads

[Patricie34]

Hi everyone,

I have a question related to the Dorado basecalling algorithm.
When performing simplex basecalling, I first convert FAST5 files to POD5 format.
Then, I use all POD5 files—fail, pass, and skip—for the downstream Dorado command.
Is it okay to do it this way, or should I use only the "pass" reads? Or does Dorado recognize this automatically?

Thank you.

Best regards,

Patricie

[malton-ont]

Hi @Patricie34,

Dorado makes no distinction regarding the source of the pod5 files - if you pass them in, they'll get basecalled just the same. Whether this is what you want to do is largely up to you.

[Patricie34]

Hi @malton-ont ,

thank you for the explanation.

And what is the best practice? Should I use only the "pass" reads, or include all of them to avoid losing important data and filter out low-quality ones after basecalling?

I am using these basecalling models: [email protected] and for some samples also [email protected]. The basecalled data are then used for variant calling.

Best regards,

Patricie

[malton-ont]

Once again, that decision is up to you and your protocol. If you would like a discussion with other bioinformaticians regarding best practices, I would suggest posting on the Nanopore community forums - this issue tracker is really for technical problems with the dorado software.

[Kirk3gaard]

Hi Patricie

Just rebasecall only your "pass" reads. The current "Q20" chemistry should return most of your data well above the Q10 cutoff for the super accuracy basecalling. So the little data that was already determined to be of poor quality is better left out. Alternatively you can of course always filter the data afterwards using e.g. https://github.com/rrwick/Filtlong

Best regards
Rasmus