Antibody staining intensity variations


I am a new user of cell profiler and have found it quite useful already. I have some questions about improving my pipeline for images with infected cells stained by an antibody to a viral protein expressed in the ER and on the surface. I have used the pipeline to identify nuclei as primary objects than postive cells as secondary objects followed by measuring and filtering for cells above a certain intensity. I can do this reasonably well for one image, but the problem I have is when I try to do a whole set of images. The intensity of the background varies from image to image as does the intensity of the positive cells. I thought there might be a way to adapt or correct for this but I’m not sure how to go about doing it.

I’ve attached some sample images and the pipeline I am currently using.


cell counting ab.cp (12.9 KB)

Here are the nuclei images to go with the stained cell images.

Hi Katie,

I’m attaching a pipeline that seems to work better though it may require some more tweaking. The main changes were:

  • Illumination correction as a pre-processing step.
  • Identifying the Ab stains as separate objects which are then related back to the cells using a RelateObjects module to establish the stains (if present) as “children” to the “parent” cells. The cells can then be filtered on the basis of the number of children, i.e, Ab-positive cells have > 1 child object.
  • Using the cell body to mask the Ab stain so that thresholding is limited to just those pixels within the cell.
  • Use of the Background threhsolding method to identify the stain, which seems to give proper results for negative vs. positive images. You may need to tweak the threshold correction factor further to make sure it works for this and other images.

Also, since detection of the cell body is dependent on the non-specific Ab stain in the cytoplasm, which is already dim and hence may not be reliable. You may want to consider using another stain for the cell body itself and use IdentifySecondaryObjects on that (e.g., phalloidin).

2011_01_05.cp (15.8 KB)

Thanks. It seems to be working better.