Why is the blocklisted feature list only for Nuclei_* ?

[dlogan]

It seems to me that the analogous Cell* and Cytoplasm* features would also have the same issues. Of course, the specific compartments would depend upon which were measured in the particular pipeline. But in general, I doubt that across the channels, for the mostly Correlation and Granularity features, that there would be any biology specific to nuclei that wouldn't also apply to the Cytoplasm and Cell compartments. @gwaygenomics

[gwaybio]

Interesting note!! We derived that blocklisted feature list using features that have previously proven unreliable - maybe the other compartments are somehow more reliable?

Either way, a more in depth study on feature reliability is probably important!

[dlogan]

We're looking at this internally on non-JUMP, but similar data, and I can report in general when that's done.

[gwaybio]

that would be great. Once you determine this, we can make version 2 of: https://figshare.com/articles/dataset/Blacklist_Features_-_Cell_Profiler/10255811 and/or update/modify the blocklist features in this repo.

Happy to also discuss authorship on that resource at a later date.

Either way, I think documenting your approach/method will be important for version 2. I am generally unsatisfied by my current lack of detailed documentation in version 1 ☹️

[AnneCarpenter]

Can anyone report what features are captured by this filter (and which aren't, that would be expected to?)

[dlogan]

@AnneCarpenter They're listed in this link: https://figshare.com/articles/dataset/Blacklist_Features_-_Cell_Profiler/10255811

Basically, various Nuclei[Correlation|Granularity] features. But not Cell[Correlation|Granularity] or Cytoplasm[Correlation|Granularity]*