Cell cycle




I’m trying to use CellProfiler to determine DNA content and fluorescence signal (for my cell cycle protein) in each cell to see if the levels of the protein fluctuate during the cell cycle. Before trying with my cells I wanted to be able to get a multi-image set from human cells that would give me a decent histogram (with defined 2N and 4N peaks). I used 4 DAPI images from the exampleSBSimage set images and tried to correct for the image to image illumination and staining variation outlined in your BMC bioinformatics paper by taking the mode value of log(integrated DNA intensity), X, for each image, converting it to log2 (by multiplying X by 3.3219), raising 2 to the power (X*3.3219)-1] to get a denominator Z, that should, when the entire corresponding image is divided by it, and a second histogram generated give me a log(integrated DNA intensity) of 0.301 (would give 1 if log base 2). However that’s not what I get, there is still considerable variation between the modes. What am I doing wrong and what do I have to correct?
I get the mode from a cellprofiler generated histogram (pipeline test, then use ratio for the histogram, and 50 bins) and then carry out the calculations using excel, then adjust the images using a second pipeline and then rerun the original pipeline. I haven’t used cell profiler analyst, since my data set is very small (200 images and only 2 parameters), and there is no one in the department with expertise to set up the database required to run cell profiler analyst. A second question is: is there a way to automatically extract the mode value of a histogram for each image in cell profiler?
Thanks a lot for your help
testPIPE.mat (950 Bytes)


In the BMC Bioinformatics paper, I’m guessing you’re referring to this paragraph in the methods section:

Then, to normalize for illumination and staining varia-
tion between slides and between images, the DNA content
measurements were log2-transformed and shifted so that
the mode of the DNA content for each image (calculated
by binning the log2-transformed DNA into 50 bins) was
equal to 1. Based on this normalized log2(DNA inten-
sity), cells were then counted as 2N, 4N, and 8N as fol-
-0.5, 0.5) was categorized as “2N"
0.5, 1.5) was categorized as " 4N”

I agree, with small image sets setting up a DB can seem like overkill especially if you don’t have someone skilled in DB management to help you. As long as your data set remains small, you should be able to look at your data and generate plots in Excel. What you want to do is take your per-object (nuclei) data and plot log2(Integrated DNA Intensity) for each image. If there is significant variation between the modes of the histograms, that’s when you have have to shift the measurements so that all the modes align and the cutoffs you set for 2N, 4N etc are consistent between images. But try looking at a few histograms because unless staining or illumination variation is very bad, the peaks should align fairly well.



…in case that wasn’t clear, here’s an image showing what I mean. These are log2(integrated DNA intensity) histograms from two different images, from two different plates actually (but same cell type and treatment, and theyve been corrected for illumination).

The blue data are from image 1, and the green data from image 2. The modes don’t align exactly, so I shifted the green data to the left, giving you the red plot. That way when you set a cutoff of what’s 2N and 4N, it would be exactly the same across images/plates.



Thank you very much Kate. Yes, I was looking for a way to avoid having to deal with the 200+ image columns in excel which will be tedious, but this method does appear to work (see attached images); after corrections although less smooth, the two broad peaks are about 2N and 4N. Well again, thanks for the prompt reply,