dorado About model complex

Hello,

I want to detect 6mA, 5mC, and 5hmC for R10 DNA with Dorado, so I need to use three models , using the Automatic model selection complex mentioned in the documentation.

However, I found that unlike the basecall of single model that can specify a model directory, this method requires three models in the current directory, otherwise it will be redownloaded . Is there a method to specify directories for multiple models ? It is indeed possible to temporarily resolve the issue through soft links, but I would still like to inquire if there is a better way to handle it.

Mar 11 '24 10:03 Yang990-sys

Hi @Yang990-sys, There's currently no way to set a search directory for automatic model selection.

I'll discuss this internally to see if we should add this as a feature as it's come up a couple of times now. I'll get back to you with an answer on this soon.

Kind regards, Rich

Mar 11 '24 11:03 HalfPhoton

We've decided to add this as a feature.

Mar 11 '24 19:03 HalfPhoton

FFR https://github.com/nanoporetech/dorado/issues/685

Mar 14 '24 14:03 HalfPhoton

@Yang990-sys,

If you'd prefer use previously downloaded models for modbase calling you can specify --modified-bases-models <path_to_model_1>,<path_to_model_2> as well as the full path to the basecalling model (i.e., don't use the automatic model detection).

Note that you actually only want 2 models here rather than 3 - the 5hmC_5mC detection is a single model, and then you can add in the 6mA as well.

Mar 25 '24 09:03 malton-ont

Hi @Yang990-sys The --models-directory option has been added to dorado-0.8.0.

Thanks for the feature request!

Kind regards, Rich

Sep 17 '24 09:09 HalfPhoton