A statistical ensemble approach to immune discrimination

The immune system needs to distinguish molecular signatures of pathogens from those found in the organisms' own proteins. A naive, but universal way to discriminate is to whitelist everything that should not elicit a reaction. Can the immune system do better? To begin to answer this question we characterize the self and pathogen proteomes as statistical ensembles. Probabilistic models reveal how both universal and phyla-specific constraints on protein evolution shape the statistics of the proteomes. The models furthermore allow us to quantify to what extent the ensembles differ systematically. We analyze whether and how these differences might be used for efficient immune defense. Finally, we compare predictions to what is known about epitopes recognized by the immune system.