jorisbalc

Thanks for the reply I'm using both a finetuned [email protected] canonical basecall model (finetuned in bonito) which manages to call the modified bases correctly (with even greater accuracy due to...

Hi @ArtRand , seems to be performing better, returning the desired result. Seemingly the GCNC find with consists of >95% GCGC targets. Though I was wondering how is it even...

Thanks for the comment, @marcus1487! I’ve noticed the lower model efficiency across different samples as well, though it’s not too bad for the task at hand. I’ll post some of...

I recently just noticed that even though in my new data with coverage of ~20X, that while evaluating the G[m]NC motif, none of the non GC[G]C kmers have any counts...

I've concluded that most of the variable motifs are due to the model efficiency, due to it being trained on fully modified ground truth data. This is, as you mentioned,...

Thanks for the incredibly fast response Marcus! As I understand further analysis requires signal refinement. I will take a look at that next. As for the read, if the spoiler...

Hi marcus, Thanks for the tips! There are a few more things I'm curious about. As far as I understand `do_fix_gauge` normalization is very strict and limits the metrics that...