Models for proteintranslocation

Hello,

I’m working with cellprofiler to identify the proteintranslocation between cytoplasm and nucleus. Right now I have 3 different models and I want to know if there could be some other, maybe better models. For all experiments I used epithelial-cells with nuclei and antibody staining with 60x magnification and I want to measure & process the IntegratedIntensity.

My first simple model is the standard propagation model which is explained at the CellProfiler page in detail.
In a second, more complex, model I included the IlluminationCorrection feature to improve the cellshapes.
The last one only includes the Distance-N module for second identification twice done, because I found that these “donut-method” is often done in the industry to get quick results.

So maybe one of you know another method that fit in here.
Anywhere, is there a common method to evaluate all the models to find the best one ? (beside RSS, AIC)

Regards,

mwolf

P.S.: Thanks for this programm =)

Hi,

You could take the approach described in our original CellProfiler paper (genomebiology.com/2006/7/10/R100). There, we quantify a translocation assay (under “Cytological profiling to reveal pathways targeted by drugs”) using the Z’-factor to evaluate the most effective measurement from large set, based on positive and negative controls. With this approach, you can evaluate your different secondary methods and see which one gives you the largest Z’-factor. CalculateStatistics can assist with this task; see the help for that module, or you can calculate it yourself in Excel.

The important thing is that you’ll need to a fair number of controls to have good statistical power since the Z’-factor is usually evaluated on a per-image basis (the per-object measures are averaged to yield a per-image value). If your assay is robust and you have a phenotype with high pentrance, you could probably use the per-object measurements to get the same thing.

Regards,
-Mark