dorado icon indicating copy to clipboard operation
dorado copied to clipboard

--modified-bases all option

Open Kirk3gaard opened this issue 1 year ago • 5 comments

Hi

I think for an improved user experience it would be great to have an option to activate "all" modifications without manually specifying the models.

Another thing I noticed is that it is confusing that --modified-bases takes a space separated list whereas --modified-bases-models takes a comma separated list. Probably easier to stick with one way of supplying arguments.

Best regards Rasmus

Kirk3gaard avatar Jun 07 '23 08:06 Kirk3gaard

Hey @Kirk3gaard

What do you imagine all doing with all-context models vs context models, i.e.5mC and 5mCG for example?

Good point on the consistency of the argument separators.

iiSeymour avatar Jun 07 '23 09:06 iiSeymour

Hi @iiSeymour

Detecting all modifications that you have models for. I think for most users whether they want to call modifications or not is a yes/no question rather than I only want to call 6mA or 5mC.

Best regards Rasmus

Kirk3gaard avatar Jun 07 '23 09:06 Kirk3gaard

Good point of Rasmus

gwl2 avatar Jun 13 '23 13:06 gwl2

I second the feature mentioned by @Kirk3gaard

amodrek avatar Jun 15 '23 13:06 amodrek

Interesting, but a couple of thoughts

  1. wouldn't specifying many models increase runtime of mod basecalling, or increase GPU RAM requirements?
  2. what about displaying mod bases as bigwig in downstream applications ? How is this handled, eg by modkit ?
  3. I guess most use cases and the best models are for 5mC

In general, I'd prefer to specify exactly which models I want unless all of the above points are solved or not detrimental.

colindaven avatar Jun 21 '23 11:06 colindaven