Search the CONFOCAL archive at
http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

 
Julio suggests that
'In my opinion, this is much more informative than a Pearson's coefficient of, let's say, + 0.43.'
 
the problem here is that the term colocalisation has a very broad meaning, covering (a) fluorophores in the same place and (b) a relationship between intensities/concentrations (Pearson correlation coefficient).
 
Being is the same place may be trivial  - the same physico-chemical properties.
A major problem with measurements that depend solely upon intensity thresholding is that they are very sensitive to the threshold chosen and in noisy fluorescent images this can be hard to set - inevitably some pixels with high noise are included while excluding pixels with a combination of low nosie and low levels of fluorescence - we can't tell then appart.
 
A relationship  between the intensities (correlation) suggests some form of interaction, this could be direct or involving an additional molecule or structure. A much potentially more interesting observation. Equally important a correlation is a sensitive measurement and likely to change with a cell's physiological state.
 
Object based methods merely pose the 'in the same place question' but can be made more interesting if a relationship between the intensities across the population of objects is tested for, or even across all pixels deemed to be in all the objects. Essentially measure the correlation in a ROI (Region of Interest).
 
While I have not seen the latest Zinchuk paper, an earlier paper (Acta Histochem Cytochem 2007, 4, 101-111) on the same topic has some has some very odd features. They very decently include scattergrams of their paired images but the scattergrams suggest that many of the pixels are saturated and therefore unsuitable for the Pearson correlation analysis that they used.  They reported Pearson correlation coefficients very close to 1, meaning a perfect match between the 2 fluorescent images, however this is not at all apparent in the accompanying scattergrams which should therefore have a single noise free straight line. We assume that the measurements come from a subset of the pixels in scattergram, but on what basis was the subset chosen ?. We twice approached the authors for clarification but none was forthcoming.
 
We have touched on some of these topics in an article in Microscopy and Analysis (September, 2008, 7-11) which should out soon.
 
Jeremy Adler
Cell Biology
The Wenner-Gren Inst.
Arrhenius Laboratories F5
Stockholm University
Stockholm 106 91
Sweden

________________________________

From: Confocal Microscopy List on behalf of Julio Vazquez
Sent: Wed 03-Sep-08 19:59
To: [log in to unmask]
Subject: Re: colocalization analysis


Search the CONFOCAL archive at http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal Judy,  

Just a few more thoughts:

There have been a number of papers on colocalization. I just found a recent one in Current Protocols in Cell Biology that gives a working protocol so to say for doing colocalization analysis with the ImageJ colocalization plugins:

Zinchuck and Zinchuk, Quantitative Colocalization Analysis of Confocal Fluorescence Microscopy Images. CPCB 4.19.1-4.19.16, June 2008. 

This can be obtained from the Journal's web site.


The various colocalization coefficients are derived from the Correlation coefficients used in statistics.  Wikipedia defines correlation as a measure of the relationship between two variables. Maybe the various correlation coefficients could be modified for use with three variables, but I am not even sure how the numbers could be interpreted, since different combinations of intensities could probably give similar results, and a lot of potentially useful information would be lost. So a workaround when using this type of approach would be to do pairwise colocalization comparisons (A vs B, B vs C and C vs A). 

If my concern were to evaluate the presence of three proteins (or other markers) inside a cell, my approach would be to select suitable thresholds for each signal, generate a mask for each channel with the selected thresholds, and then combine the three masks through Boolean or Image Arithmetic operations. This would generate a channel where all three markers are present (above set threshold). You could also create masks by combining two markers, and use that mask to analyze the distribution of the third marker. I would then quantitate the different markers through those masks (integrated intensity, surface area/volume, or whatever variable is the most relevant biologically), and either look at the absolute numbers, or the relative numbers as a function of the total cell area, or total intensity in any given channel, etc... To be even more accurate, I could make a mask for the cells (or nuclei, etc...) and restrict my analysis to those regions. As I mentioned in the previous post, it should be up to the investigator to let you know which of these numbers or pairwise comparisons are the most relevant biologically for their experiment. With the mask approach you can determine, for instance, that 20% of protein A is present in regions where proteins B and C are also present, and so on... this is very easy for me to grasp intuitively. I will also be able to know how large the colocalized patches are, how many there are, where they are located, and so on... In my opinion, this is much more informative that a Pearson's coefficient of, let's say, + 0.43. 

My main issue with the standard correlation approach is that it is pixel-based, as opposed to object-based, and that the results are highly dependent on how the images were collected (sampling) and processed, how much noise and background they have, how well registered the different channels are (always true, but even more so in this case, especially if looking at colocalization of very small objects) and which regions and thresholds are used for analysis. Used properly, these methods are quite powerful though, and will give very useful information such as colocalization coefficients,  % of A colocalized, % of B colocalized, etc....  but I have seen so many results that didn't make sense just because someone was using a poorly sampled image, or a noisy image (such as a confocal image under limiting signal conditions), that I am very careful and tend to prefer a more visual approach where objects are identified based on thresholds, and then analyzed.  Anyhow, the paper above discusses some of the issues to take into consideration when performing the pixel-based type of colocalization analysis.

hope this helps, 

with kind regards, 

Julio.


--
Julio Vazquez
Fred Hutchinson Cancer Research Center
Seattle, WA 98109-1024

http://www.fhcrc.org <http://www.fhcrc.org/> /

-

On Aug 27, 2008, at 6:58 AM, Judy Trogadis wrote:


	Search the CONFOCAL archive at
	http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal

	Hi, Michael,

	You're right, 2 runs would do it - but one of the users of our imaging facility has triple labelled preps ready to analyze. He is looking for the presence of 3 proteins in a cell but I am not sure about their proximity to each other. Visual observation at high magnification could give a clue. I think the user wants some numerical value for a grant. 

	Thanks.
	Judy



				Michael Weber <[log in to unmask]> 08/27/08 9:45 AM >>>

	Search the CONFOCAL archive at
	http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal 

	Judy,

	colocalization analysis is quite straight forward with ImageJ and the
	"Colocalisation Threshold" plus "Colocalization Test" plugin according to
	Costes.

	Regarding triple labeling, which type of questions do you plan to answer?
	I can think about a scenario with two marked structures and how they are
	colocalizing with the nuclei - then you do two runs: nuclei vs. staining
	1, nuclei vs. staining 2. I am not aware of an established three-color
	colocalization equation. Or do I miss something here?

	Michael



		Search the CONFOCAL archive at
		http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal 

		I have a triple labelled sample and would like to do colocalization
		analysis. What approaches are most people using? The plugins I have seen
		or used only handle double labelled specimen. A 3-D fluorogram perhaps?

		Thank you.

		Judy Trogadis
		Bio-Imaging Coordinator
		St. Michael's Hospital, 7Queen
		30 Bond St.
		Toronto, ON M5B 1W8, Canada
		ph:  416-864-6060  x6337
		pager: 416-685-9219
		fax: 416-864-6043
		[log in to unmask]