Deciphering Interactions in Causal Networks without Parametric Assumptions

Abstract

With the assumption that the effect is a mathematical function of the cause in a causal relationship, FunChisq, a chi-square test defined on a non-parametric representation of interactions, infers network topology considering both interaction directionality and nonlinearity. Here we show that both experimental and in silico biological network data suggest the importance of directionality as evidence for causality. Counter-intuitively, patterns in those interactions effectively revealed by FunChisq enlist an experimental design principle essential to network inference -- perturbations to a biological system shall make it transits between linear and nonlinear working zones, instead of operating only in a linear working zone.