Re: colocalization - three channels

[Daniel James White]

Dear Judy,

I just had a look at the Colocalize RGB plugin,
and it looks like it does a good job of a simple colocalisation  
analysis,
but doesn't give you any statistics, so the interpretation of the  
results is pretty subjective.

There is an auto threshold method, but from the code snippet below,
I cant see how it works.... maybe its using an internal imageJ method?  
Stephan? any clues?
// auto-threshold each image if desired
        if (auto) {
                ImageProcessor ip;
                for (int i = 0; i < image.length; i++){
                        for (int j = 0; j < zdepth[i]; j++) {
                                pixels = new byte[ (width[i]*height[i]) ];
                                ip = new ByteProcessor(width[i], height[i]);
                                ip.setPixels(pixels);
                                thresh[i] = ip.getAutoThreshold();
                        }
                }
        }
        return true;

A very good way to auto threshold is to use the method of costes et al ,
which uses pearsons correlation coefficients above and below the  
thresholds to determine
the thresholds for a pair of channels, where there is 0 correlation  
between the channels below the threshold.
This is objective and repeatable, and mathematically sensible.
This is implemented in the colocalisation threshold plugin.

Automatic and Quantitative Measurement of Protein-Protein
Colocalization in Live Cells
Sylvain V. Costes,*y Dirk Daelemans,z Edward H. Cho,* Zachary Dobbin,*  
George Pavlakis,z
and Stephen Lockett* Biophysical Journal Volume 86 June 2004 3993–4003  
3993

However, this doesnt work out of the box for 3 channels and you need  
to do the individual pairs one by one.
In the same paper the Costes described this auto threshold (I will  
send it to you)
there is also a statistical significance test described,
which is implemented in the
Colocalisation Test plugin.
This is super important as it tells you if the coloc you see is random  
or real.
For a busy image it could easily be random.

Further, 2D histograms displaying the spatial intensity correlation of  
2 channels
are a much better way to display colocalisation than colour merge  
images (which are usually very misleading
since our brains always look for high contrast and miss most of the  
information that is there)

All these methods could be extended to n channels,
but the computation would get very long, and it is difficult to  
display or print a 3D hisotgram..

My colleage Stephan Preibisch suggested a better way of measuring 3  
channel colocalisation using an
entropy method..... maybe now ids a good time to try to implement that  
as an imageJ plugin.... Stephan?

cheers

Dan








On Aug 28, 2008, at 6:00 AM, IMAGEJ automatic digest system wrote:

> Date:    Wed, 27 Aug 2008 22:16:12 +0900
> From:    eiji <[hidden email]>
> Subject: Re: colocalization
>
> Dear Judy,
> I think Colocalize RGB plugin could analyze triple labelled samples.
> Good luck.
>
>
> --
> Eiji SHINYA, M.D., Ph.D., FJSIM.
> Senior Assistant Professor
> Nippon Medical School
> Department of Microbiology and Immunology
> 1-1-5 Sendagi
> Bunkyo city
> Tokyo, 113-8602
> JAPAN
> Tel.           [phone] ext. 5547
> Skype In   [phone]
> Fax.          [phone]
> E-mail       [hidden email]
> Skype       eshinya

Dr. Daniel James White BSc. (Hons.) PhD
Senior Microscopist / Image Processing and Analysis
Light Microscopy Facility
Max Planck Institute of Molecular Cell Biology and Genetics
Pfotenhauerstrasse 108
01307 DRESDEN
Germany


New Mobile Number!!!

[phone] (German Mobile)
+49  (0)351 210 2627 (Work phone at MPI-CBG)
+49  (0)351 210 1078 (Fax MPI-CBG LMF)

http://www.bioimagexd.net
http://www.chalkie.org.uk
[hidden email]
( [hidden email] )