Subjective Colocalization Analysis with Fuzzy Predicates

Understanding the protein-to-protein interactions at the subcellular level, as well as other organic molecules, is crucial to explain cellular functions and to elucidate disease mechanisms. These interactions can be captured visually by overlapping the fluorescent microscopic images of two proteins tagged with fluorescent labeling agents that react to green and red wavelengths respectively. Interaction is determined by subjectively assessing the amount colocalization of green and red on the image composite based on the amount of yellow present in the image composite (i.e., green and red form yellow). Attempts to reduce the subjectivity of this process have focused on the computation of statistical coefficients and related methods. Even though statistical colocalization coefficients give a degree of correlation among the imaged proteins, they still need to be interpreted with subjective qualifiers like ”high”, ”low”, ”strong”, etc. Hence, there is no current agreement on the meaning of these coefficients among researchers. In this paper we propose the use of fuzzy linguistic variables to model the subjective interpretation of co-localization coefficients. Based on interpretations found in the literature, we produce a set of rules that map the coefficient values to a linguistic interpretation. The result of this work is a tool that provides an descriptive ensemble of coefficient interpretations that could guide researchers to a uniform interpretation colocalization criteria.