High throughput experiments, characteristic of studies in systems biology, produce large output data sets often at different time points or under a variety of related conditions or for different patients. In several recent papers the data is modeled by using a distribution of maximal information-theoretic entropy. We pose the question: ‘whose entropy’ meaning how do we select the variables whose distribution should be compared to that of maximal entropy. The point is that different choices can lead to different answers. Due to the technological advances that allow for the system-wide measurement of hundreds to thousands of events from biological samples, addressing this question is now part of the analysis of systems biology datasets. The analysis of the extent of phosphorylation in reference to the transformation potency of Bcr-Abl fusion oncogene mutants is used as a biological example. The approach taken seeks to use entropy not simply as a statistical measure of dispersion but as a physical, thermodynamic, state function. This highlights the dilemma of what are the variables that describe the state of the signaling network. Is what matters Boolean, spin-like, variables that specify whether a particular phosphorylation site is or is not actually phosphorylated. Or does the actual extent of phosphorylation matter. Last but not least is the possibility that in a signaling network some few specific phosphorylation sites are the key to the signal transduction even though these sites are not at any time abundantly phosphorylated in an absolute sense.
Information theory Prior distribution Systems biology Signal transduction High throughput experiments Phosphoproteomics