Genetic interactions from first principles
We derive a general statistical model of interactions, starting from probabilistic principles and elementary requirements. Prevailing interaction models in biomedical researches diverge both mathematically and practically. In particular, genetic interaction inquiries are formulated without an obvious mathematical unity. Our model reveals theoretical properties unnoticed so far, particularly valuable for genetic interaction mapping, where mechanistic details are mostly unknown, distribution of gene variants differ between populations, and genetic susceptibilities are spuriously propagated by linkage disequilibrium. When applied to data of the largest interaction mapping experiment on Saccharomyces Cerevisiae to date, our results imply less aversion to positive interactions, detection of well-documented hubs and partial remapping of functional regions of the currently known genetic interaction landscape. Assessment of divergent annotations across functional categories further suggests that positive interactions have a more important role on ribosome biogenesis than previously realized. The unity of arguments elaborated here enables the analysis of dissimilar interaction models and experimental data with a common framework.