On the pixel selection criterion for the calculation of the Pearson's correlation coefficient in fluorescence microscopy

Abstract

Colocalisation microscopy analysis provides an intuitive and straightforward way of determining if two biomolecules occupy the same diffraction-limited volume. A popular colocalisation coefficient, the Pearson's correlation coefficient (PCC), can be calculated using different pixel selection criteria: PCC_ALL includes all image pixels, PCC_OR only pixels exceeding the intensity thresholds for either one of the detection channels, and PCC_AND only pixels exceeding the intensity thresholds for both detection channels. Our results show that PCC_ALL depends on the foreground to background ratio, producing values influenced by factors unrelated to biomolecular association. PCC_AND focuses on areas with the highest intensities in both channels, which allows it to detect low levels of colocalisation, but makes it inappropriate for evaluating spatial cooccurrence between the signals. PCC_OR produces values influenced both by signal proportionality and spatial cooccurrence but can sometimes overemphasise the lack of the latter. Overall, PCC_AND excels at detecting low levels of colocalisation, PCC_OR provides a balanced quantification of signal proportionality and spatial coincidence, and PCC_ALL risks misinterpretation yet avoids segmentation challenges. Awareness of their distinct properties should inform their appropriate application with the aim of accurately representing the underlying biology.