Quantitative determination of spatial protein-protein correlations in fluorescence confocal microscopy

To quantify spatial protein-protein proximity (colocalization) in paired microscopic images of two sets of proteins labeled by distinct fluorophores, we showed that the cross-correlation and the autocorrelation functions of image intensity consisted of fast and slowly decaying components. The fast component resulted from clusters of proteins specifically labeled, and the slow component resulted from image heterogeneity and a broadly-distributed background. To better evaluate spatial proximity between the two specifically labeled proteins, we extracted the fast-decaying component by fitting the sharp peak in correlation functions to a Gaussian function, which was then used to obtain protein-protein proximity index and the Pearson's correlation coefficient. We also employed the median-filter method as a universal approach for background reduction to minimize nonspecific fluorescence. We illustrated our method by analyzing computer-simulated images and biological images.