Automated interpretation of subcellular location patterns

Fluorescence microscopy is widely used to analyze the distribution of proteins within cells. As currently practiced, the assignment of a protein to a particular organelle is done by visual inspection of images or comparison between the distribution of the unknown protein and markers with known location patterns. In order to use fluorescence microscopy for large scale or proteome-wide analysis of protein location, improved approaches that are more automated, objective, and sensitive are needed. Our group has therefore developed automated systems that can recognize the major subcellular location patterns in images of single cultured cells, and has shown that these systems are more sensitive than visual inspection. The foundations of these systems are sets of numerical features that describe the essential characteristics of a subcellular pattern without being overly sensitive to the size, shape and orientation of that cell within the field of view. These features can be used to measure the similarity between protein patterns and therefore for the first time to group proteins in an objective manner based on their high-resolution patterns.