Home

Colocalization Colormap

Colocalization studies are widely performed using methods based on global image analysis which for instance involves intensity correlation coefficients such as the Pearson correlation coefficient and the Manders overlap coefficient The general paradigm of these approaches is to plot the pixel values of two images against each other and display the result in a pixel distribution diagram called a scatter plot in a way in which an intensity of the given pixel in the first channel is used as the x-coordinate of the scatter plot and the intensity of the corresponding pixel in the second channel consists its y-coordinate. A linear, or non-linear relationship among the intensities of both channels might be then assumed. The disadvantage of methods based on general intensity correlation is lack of information about the spatial distribution of colocalized signals with respect to the composite nature of biological structures within the particular image. This website contains Colocalization Colormap ImageJ plugin that implements the Jaskolski's algorithm (Jaskolski et al. 2005). This innovative method produces a pseudo-color map of correlations between pairs of corresponding pixels in two original input images, thereby offering quantitative visualisation of colocalization.

Spatial representation of colocalization

In course of the analysis, according to the equation below, the plugin calculates normalized mean deviation product (nMDP) which mathematically represents correlation between intensities of corresponding pixels (values are ranging from -1 to 1).

Further the program generates image that contains spatial distribution of calculated nMDP values. The distribution is based on a color scale in which negative nMDP values are represented by cold colors (segregation). On the other hand, values above 0 are represented by hot colors (colocalization). Consecutively, this allows for creation of spatial map of colocalization as demonstreted below.

Synaptic localization of cadherin-9 at mossy fiber terminals with Colocalization Colormap

The association of cadherin-9 immunoreactivity with mossy fiber pathway was demonstrated with Colocalization Colormap (top panels). Moreover, single confocal plain of mouse brain section containing cadherin-9 immunoreactivity within mossy fiber boutons was analyzed with the plugin and nMDP distributions as well as Icorr value were calculated to assess the colocalization between cadherin-9 and mossy fiber terminal (bottom panels).

Measurement of colocalization.

Plugin calculates index of correlation (Icorr). The index represents fraction of positively correlated (colocalized) pixels in the analyzed images which allows for very sensitive quantitative measurement of colocalization.

Simulated data

The proper algorithm implementation was tested on artificial images prepared to simulate complete colocalization, partial colocalization, and lack of colocalization. For that purpose, we have created an artificial spherical object based on a 3D intensity gradient (fig. 1a). The object was represented by a stack of images, that was first duplicated and then all the pixels in the duplicated stack were translated (fig. 1b). Complete colocalization was produced with translation by 0 pixels, whereas partial colocalization was produced with translation by 10, 20, or 30 pixels, and lack of colocalization (separation of the objects) was produced with translation by 40 pixels (fig. 1c, top panels). Reduced intensity correlation between pairs of corresponding pixels within generated images was observed proportionally to fewer overlap between objects (fig. 1c, middle panels). This set of data with dissimilarly colocalized spheres was then analyzed with Colocalization Colormap to calculate nMDPs distribution and Icorr values (fig. 1c, middle panels). The threshold value for the analysis of all translated objects was set as ‘118’ due to the fact that this was the minimum threshold value that assured complete separation of objects translated by 40 pixels (fig. 1c, bottom panels). As a result, we observed gradual decrease of Icorr values that was proportional to the decrease of colocalization between analyzed images. Icorr values were in range of 1 (for complete colocalization) to 0 (for lack of colocalization) (fig. 1c, bottom panels). Intermediate Icorr values for partial colocalization were as follows: 0.573 for 10-pixel translation, 0.132 for 20-pixel translation, and 0.055 for 30-pixl translation (fig. 1c, bottom panels). Decrease of colocalization between analyzed objects was also indicated by the loss of warm colors in corresponding colocalization maps due to reduced fraction of nMDP values with indexes above zero (fig. 1c, bottom panels).

In addition, the color map of duplicated stack was also inverted in order to generate pair of stacks containing negatively correlated intensities that simulate lack of colocalization while analyzed specimens overlap (exclusion of objects) (fig 1d). In this case we used the ‘autothreshold’ option for the analysis. Calculated Icorr value equaled 0 demonstrating lack of colocalization between objects. The negative global intensity correlation between stacks was reflected by color map of nMDPs distribution that did not contain any positively correlated pairs of corresponding pixel intensities (fig 1d, panels on the right).

fig. 1 Validation of Colocalization Colormap on simulated data.

a) The plugin was tested on simulated data mimicking progressive separation of the two fluorescent foci. For that reason, stack of images containing 8-bit-encoded intensity gradient (left graph) with radial 3D distribution that renders as a sphere (middle and right panel) was created.

b) The stack was duplicated and then the duplicated set of images was translated in x and y directions by the defined number of pixels (pxl).

c) Progressive translation resulted in decrease of colocalization and intensity correlation between images (c, top and bottom panels). This set of incrementally separated artificial spheres was then analyzed with Colocalization Colormap plugin to create maps of colocalization (nMDP distribution) and to calculate the corresponding Icorr values (middle panels).

d) The plugin was also tested on the two negatively correlated signals. For that purpose the spherical gradient was duplicated and inverted. Two artificial channels created this way were analyzed with Colocalization Colormap to demonstrate lack of colocalization due to negative correlation of the intensities (right panel).

Google Sites

Report abuse

Colocalization Colormap

Spatial representation of colocalization

Measurement of colocalization.

Simulated data

Download

Documentation

Contact