endrebak
I like your summary Stephen. I think the concatenation, not the splitting, is the expensive step though. On Thursday, October 21, 2021, Endre Bakken Stovner ***@***.***> wrote: > If you...
I have also realized that subsetting on one interval can be done in pure numpy, without the NCLS. Should update pyranges and post an example here On Monday, October 25,...
If you could send me those I’d welcome them. I’d also appreciate your scripts to produce these wheels. Perhaps they can be made into a GitHub action On Friday, January...
I think it is just an overflow error. You have too many hits?
I’ll reply more in depth on monday, when I’m back at work :) Pyranges should be able to do this. The repo is on my github :)
pyranges is still largely unused. I have passing unittests, but it might still have bugs or not work. I would also look into this potential error in bedtools jaccard: https://github.com/arq5x/bedtools2/issues/645...
Also, if you use pybedtools, it is advisable to presort the data first. It is much faster then.
Here: https://github.com/biocore-ntnu/ncls/blob/master/ncls/src/ncls32.pyx#L111 Check why it does not search again when the number of intervals is > 1024 https://github.com/biocore-ntnu/pyranges/issues/220
Thanks for reporting. NCLS is not meant to be used with negative positions though. What is your use-case for that? I guess I should add a check to see that...
I completely agree and I am happy you reported it. I actually only use the vectorized functions because they are so much faster.