Colocalization analysis is the most common technique used for quantitative analysis of fluorescence microscopy images. Several metrics have been developed for measuring the colocalization of two probes, including Pearson's correlation coefficient (PCC) and Manders’ correlation coefficient (MCC). However, once measured, the meaning of these measurements can be unclear; interpreting PCC or MCC values requires the ability to evaluate the significance of a particular measurement, or the significance of the difference between two sets of measurements. In previous work, we showed how spatial autocorrelation confounds randomization techniques commonly used for statistical analysis of colocalization data. Here we use computer simulations of biological images to show that the Student's one-sample t-test can be used to test the significance of PCC or MCC measurements of colocalization, and the Student's two-sample t-test can be used to test the significance of the difference between measurements obtained under different experimental conditions.
A common technique in microscopy is to label one substance red, label another substance green, and see whether the red and green substances are found in the same areas of the image (‘colocalize’). A correlation coefficient can be used to measure the amount of colocalization, but it has been difficult to know whether the amount of colocalization was statistically significantly more than expected by chance. Here we use computer simulations to show that a good statistical technique is to measure the correlation coefficient between the red and green signals in each of several images, then use Student's t-test to see whether the average correlation coefficient is significantly greater than zero.