uLTRA + SQANTI3 about ultra HOT 2 CLOSED

Hi Fran,

Since you mention that the reads usually are processed, I think --isoseq parameter is fine. However, please skip the --ignore_rc flag. This was an experimental parameter used in development. Looking through the codebase I'm actually unsure it has any effect.

Another parameter is the number of cores/threads --t. Default is 3 and uLTRA's memory footprint grows with the number of cores, it may be good to set this to what SQANTI3 uses.

I'm not sure about the most common use case you see with SQANTI3 (e.g., isoseq or ONT, human/non-human, and the number of reads in the datasets), but that is what I could think about if you have to choose a fixed parameter setting for uLTRA.

I recently implemented an option --use_NAM_seeds there are some tradeoffs with this option - faster and smaller intermediate files, but can be more memory consuming/slightly less accurate. If you're interested, see here.

Glad that you want to try it out and happy to receive feedback if you get any!

Best,
Kristoffer

from ultra.

Thanks for your quick response and your advise! I will remove the --ignore_rc flag and set the --t parameter as you said. Regarding the --use_NAM_seeds, I read about this option, but I will keep it "off". The idea is to map transcript-models, so it shouldn't be a problem the amount of sequences. The reference transcriptome from mouse has ~140k transcripts, it would be crazy to have >5M transcripts...

I'll let you know any insight that we get of this. But I won't open an issue, but I'll contact you :)

Bests,
Fran

from ultra.