Ying-Lab / scIBD

A doublet detection tool for scCAS data
MIT License
3 stars 1 forks source link

reject detected doublets #2

Closed Genshin-Impact-king closed 1 day ago

Genshin-Impact-king commented 8 months ago

Thanks for your great work! But I have a question about if a droplet will not change once it is detected as a doublet? And what does this code mean below? Why droplets whose 'label_refer != 0.05' is not doublet? Wouldn't there be any other numbers from 0-0.1 other than 0.05. image

Ying-Lab commented 3 days ago

Hi! Sorry for the serious delay. The maintainer was in some issue in the last few months. I have updated the comments in the latest codes. The label of a droplet (doublet or others) is temporary in one iteration, which only influences the next iteration, including the clustering and the doublet score calculation. The final determination of a droplet to be a doublet must refer to all the iterations (the mean doublet scores across all iterations). For the second point, 'label_refer != 0.05' is the remained codes in one aborted version. In this version, the predicted doublets in one iteration will be discarded by a tuned rejection criterion, and will be assigned with a 0.05 label to distingush these droplets. This version has beed dropped. For that It is almost impossible for a normalized doublet score to be exactly 0.05, the codes have none matter on the final results. Best wishes Scotty Zhang

Genshin-Impact-king commented 3 days ago

Thanks for your reply