Identification of subcellular features for data extraction (Cell Painting)

I have carried out a cell painting assay (staining the nucleus, ER, mitochondria, golgi, F-actin and nucleoli using six dyes) and used the example features pipeline from the Carpenter lab, which gives a great starting point for data extraction.

So far I can identify nuclei, cells and cytoplasm and extract data on object intensity, distribution and texture etc. However, I also need to identify the ER, golgi, mitochondria and actin! I have tried using the speckles module but cannot seem to do this for golgi and mitochondria.

Has anyone had experience of defining these organelles as objects and extracting data from images of these organelles? I have attached the pipeline I have so far, any help would be really appreciated!

Feature_Extraction_FromCarpenterTemplate_031116.cpproj (1.7 MB)


When we do Cell Painting we actually do not attempt to identify individual organelle components. We found that segmentation was pretty unstable for these organelles (in part because biologically speaking they are not exactly distinct “objects” at this resolution, they are more like interconnected networks). That led to rather noisy resulting data.

All of the successful results we’ve had using Cell Painting rely on texture/intensity measurements alone to identify unusual morphology in the ER/golgi/mit/actin.

That’s not to say it isn’t worth giving it a try if you want to experiment (and especially if you have controls where you can really test if it’s helping!) One option is to try the EnhanceOrSuppressFeatures module to make the organelles of interest more distinct. Speckle enhancement is really geared for truly round objects so it won’t likely work well. The neurite option might be better, though I’m not claiming that is a great fit to those organelles either.

I’d be excited to hear whether you pursue this and give it a try, please report back if you do!