Skip to content, Skip to search

Changes

Colocalization Analysis

1,115 bytes added, 07:38, 31 July 2015
Pixel intensity spatial correlation analysis: add clarity and a little detail and link to alder2010 refuting the overlap coefficient
==== Pixel intensity spatial correlation analysis ====
There Here just are two of many colocalization coefficients to express the fraction intensity correlation of colocalizing objects in each component of a dual-color image:# '''Pearson's correlation coefficient.''' Not It is not sensitive to differences in mean signal intensities or range, or a zero offset between the two components. The result is +1 for perfect correlation, but 0 for no correlation, and -1 for perfect anti-correlation. Noise makes the value closer to 0 than it is not unambiguous because of the strong influence of the ratio of the number of objects in the two componentsshould be.# '''Manders split coefficients.''' Proportional to the amount of fluorescence of the colocalizing objects pixels or voxels in each component, which is dependent on the intensities of the signalscolour channel. You can get more details in [[Media:Manders.pdf|Manders et al.]]Values range from 0 to 1, expressing the fraction of intensity in a channel that is located in pixels where there is above zero (or threshold) intensity in the other colour channel.
These coefficients measure the significance amount or degree of true colocalization, or rather correlation and co-occurrence respectively (but should not be expressed as % values, because that is not how they are defined). The But if there is nothing to compare them to, what do they mean? A statistical significance test is was derived by Costses to evaluate the probability that the measured value of Pearson's correlation, r from between the two colors colour channels is significantly greater than values of r that would be calculated if there was only random overlapof the same information. This test is performed by randomly scrambling the blocks of pixels (instead of individual pixels, because each pixel's intensity is correlated with its neighboring pixels) in one image, and then measuring the correlation of this image with the other (unscrambled) image. You can get more details in [[Media:Costes etalColoc.pdf|Costes et al.]]The result of this tests tell us if the Pearsons r and Manders' coefficient we measure are better than pure chance or not.
Other coefficients include ranked correlations such as Spearman and Kendal's Tau, %overlap of intensities or area (volume).Some others are described in the literature that have been used in publications, but that have been refuted as insensetive, such as the overalp coefficient from the Manders paper, which [[Media:Adler_et_al-2010-Cytometry_Part_A.pdf|J. Adler et al.]] showed to be not very useful compare to Pearson's r and Mander split coefficients.  Other methods include ICCS (image cross correlation spectroscopy) and a derivative of that called PPI (protein proximity index. Maybe someone wants to add PPI to the Coloc_2 plugin? [[Media:Wu2010-crosscorrelationPPI_Coloc.pdf|original article here]]
==== Object-based overlap analysis ====