http://imagej.273.s1.nabble.com/counting-of-colocation-area-tp3696372p3696374.html
feedback / comments / criticism / objections / abuse / etc. welcome
> From: Dan White <
[hidden email]>
> Date: April 29, 2008 12:21:59 PM GMT+02:00
> To: ImageJ Interest Group <
[hidden email]>
> Cc:
[hidden email]
> Subject: Re: counting of colocation area
>
> Dear Gus,
>
> Using a 2 colour merge to look for "co-location" of proteins is a
> poor mans way of doing it.
> This simple way is very prone to error as it is totally subjective:
> The amount of overlap you see depends on the relative brightness of
> the 2 colour channel/images.
> Some people are colour blind and cant see the results (for red and
> green especially).
> Green looks brighter than red, which looks brighter than blue,
> so your brain and eyes have a hard time seeing what is really going
> on, especially for lower intensities overlapping.
> There are no statistical tests involved to determine if what you see
> is random or real,
> so what you see in a single image could be misleading.
>
> Remember the colours are false anyway, DAPI isnt 100% blue, and RFP
> isnt 100% red,
> so it is in many ways silly to represent them in a black to single
> colour R G or B
> Better to use a grey scale or spectrum look up table which our eyes
> can see better.
>
>
> Search this list for threads on colocalisation / colocalization
> and there is lots of discussion on how to do it objectively and
> quantitatively.
>
> In short,
> use the methods of Manders and Costes
> which are in the colocalisation plugins for imageJ
>
> This gives you:
> 1) Manders coefficients - which describe the proportion of one color
> colocalised with the other, and vice versa.
> These are insensitive to differences in the intensity between
> colour channels, and so robust between different images of the same
> sample and possibly even different samples.
> 2) a way to set the thresholds (what is real signal and what is
> background) in a reliable objective and reproducible way (Costes)
> 3) A statistical test to see if the colocalisation is random or real.
> 4) A 2D histogram, where you can much better see correlation in
> spatial intensity distribution than in a 2 colour merge image.
> 5) A map of the colocalised pixels to compare with the original
> images.
>
> Images must be correctly spatially sampled, according to Nyqvist
> theorem (about 80-100 nm in xy and about 250 nm in z for a 1.4 NA
> objective, so you don't miss some of the information the microscope
> could see).
>
> Fluorescent beads should be used as positive controls (try tetraspek
> from invitrogen),
> so you can see if the different colours appear in the same place.
> (even if you use apochromat objectives , they might not be,
> and you may need to precisely align your scanners/microscopes
> optics, or even correct the images in post processing. Objects at
> the edges of a large field of view suffer badly from chromatic
> aberration quite often)
>
>
> I strongly believe that pixel intensity based colocaisation analysis
> results should:
>
> 1) show both colour channels images using a spectrum look up table
> (not black to red / blue /green - the dyes are not 100% R G or B
> anyway)
> 2) show the thresholds used (as calculated by the Costes method)
> 3) Show the thresholded Manders coefficients,
> 4) show the statistical significance according to Costes, Fay or van
> Steensel)
> 5) show the 2D histogram (which is very informative)
> 6) be done in 3D in a 3D sample (else you miss much important
> information)
> 7) could show a colour merge image (i really dont like those though)
> - but better a colocalisation map
>
>
> The same methods are also implemented in BioImageXD, and it should
> be faster (also free software)
> Imaris also has the same maths - nearly. Imaris is nice, but
> expensive.
>
> You can also do object based colocalisation, but thats another
> story, involving segemtning objects out of the images. ...
>
> I will send you the pdf files for Manders and Costes off list
> (anyone else want them... Manders is hard to find)
> Read, digest, understand, become enlightened. The concepts are quite
> simple, don't be scared off by the equations.
>
> cheers
>
> Dan
>
>
>
> On Apr 29, 2008, at 6:00 AM, IMAGEJ automatic digest system wrote:
>
>> Date: Mon, 28 Apr 2008 02:05:54 -0700
>> From: agus wijoyo <
[hidden email]>
>> Subject: counting of colocation area
>>
>> Dear Forum
>>
>> I would like to find out the location of my protein in the cells,
>> the prote=
>> in was stained with Red Fluorescent Protein, and located in both
>> the cytopl=
>> asm and nucleus. The nucleus of the cells was stained with DAPI
>> (blue colou=
>> r). After merge the 2 image, there were part of protein which was
>> located i=
>> n the nucleus, and make colocation with the blue colour of the
>> nucleus. I w=
>> ant to count the percentage of the protein in the nucleus area and
>> in the c=
>> ytoplasm. Please give me an advise
>>
>> =20
>>
>> Thank you=20
>>
>> Gus
>> =20
>
> 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!!!
>
> +49 (0)15114966933 (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] )
>
>
>
>
>
Dr. Daniel James White BSc. (Hons.) PhD