Testing and overcoming the limitations of Modular Response Analysis

Abstract

Modular Response Analysis (MRA) is an effective method to infer biological networks from perturbation data. However, it has several limitations, such as strong sensitivity to noise, need of performing independent perturbations that hit a single node at a time, and linear approximation of dependencies within the network. Previously, we addressed MRA sensitivity to noise by revisiting MRA as a multilinear regression problem. Here, we provide new contributions to complement this theory. First, we overcame the need of perturbations to be independent, thereby augmenting MRA applicability. Second, using analysis of variance (ANOVA) and lack of fit tests, we assessed MRA compatibility with the data and identified the primary source of errors. If nonlinearity prevails, we propose a polynomial extension to the model. Third, we demonstrated how to effectively use the prior knowledge of the network studied. Finally, we added these innovations to our R software package MRARegress to provide a complete, extended theory around MRA and to facilitate its access by the community.