4 color analysis

Hello,

I find CellProfiler very helpful in quantifying the intensity of two cytoplasmic proteins, and would like to analyze their correlation with another nuclear factor, yet this requires the use of four color analysis.

To identify the nucleus I used DAPI (blue) and the other two proteins are stained in green and red, the fourth protein was stained with Cy5 and I have pseudo-colored it in white. My question is how can preform four color analysis using CellProfiler? when I created four image files (one for each color) I was not able to use all four and only red, blue and green.

Thanks a lot, :smiley:
Roy

Hi Roy,

Could you post an example set of the four image files?
-Mark

Hi Mark,

Here is an example of the pictures I am analyzing. I wish to examine the levels of the nuclear factor in cells that express both cytoplasmic proteins; so far I was able to detect the double positive cells and now want to examine the expression of the nucleic factor.

Thanks,

Roy


Hi Roy,

Sorry, I should have been more specific: Could you post the raw image files for the 4 channels for the example in your figure, plus the pipeline that you are currently using?
-Mark

Hi Mark,

Sorry for the misunderstanding… here are the files, hope I got it this time :smile:

(Please note that currently my pipeline contains only 3 colors)

Thanks a lot!
Roy
Roy pipeline (tumors)_3.cp (8.35 KB)








Hi Roy,

My suggestion is to find the nuclei, the red stain and the green stain as 3 separate primary objects, rather than taking the primary/secondary approach. The reason is that every primary object receives a secondary object, whether the stain is positive or not, and then the challenge becomes figuring out which secondary objects represent positive cell stains.

Instead, you can identity each stain as an object, and then use the RelateObjects module to establish a parent-child relationship between the nuclei (parent) and the stains that they overlap with (child). If a nuclei overlaps with a stained object, it receives a non-zero child count. You can then use FilterObjects to remove nuclei that have red stain child count of 0 and a green stain child count of 0. These remaining objects can be used to measure the intensity from the white channel.

I’m attaching a pipeline which does what I’ve described above. You may need to do some tweaking to make that the regions you consider positive in the red and green channels are being identified properly, as well as making sure that the nuclei are identified well.

Hope this helps!
-Mark
2011_12_05.cp (11.3 KB)

Hi Mark,

Thanks a lot for your great assistance! I’ll try this out a bit and post my findings…

Roy

After running the pipeline a few times it looks to be working well (with minor tweaking). I now want to compare the double positive cells to the single positive (for each dye) and see if there is a difference; I was wondering what would be the best way to do that.

I thought about creating a new filterObjects and tell it take each of the single positive objects (by marking 1 in min value) and then not taking the double positives by marking it’s min 0. Is that a methodology correct ? is there a better way to do accomplish this?

Thanks,
Roy

I you’re on the right track. By definition, a single-positive object is one in which the number of children is at least 1 in the stain of interest, and exactly 0 in the others. You will need to do this for each of the stains of interest, of course.
-Mark