A logic-based method to build signaling networks and propose experimental plans

Adrien Rougny,Nathalie Langonné,Eric Reiter,Pascale Crépieux,Christine Froidevaux,Anne Poupon,Pauline Gloaguen

doi:10.1038/s41598-018-26006-2

Abstract

With the dramatic increase of the diversity and the sheer quantity of biological data generated, the construction of comprehensive signaling networks that include precise mechanisms cannot be carried out manually anymore. In this context, we propose a logic-based method that allows building large signaling networks automatically. Our method is based on a set of expert rules that make explicit the reasoning made by biologists when interpreting experimental results coming from a wide variety of experiment types. These rules allow formulating all the conclusions that can be inferred from a set of experimental results, and thus building all the possible networks that explain these results. Moreover, given an hypothesis, our system proposes experimental plans to carry out in order to validate or invalidate it. To evaluate the performance of our method, we applied our framework to the reconstruction of the FSHR-induced and the EGFR-induced signaling networks. The FSHR is known to induce the transactivation of the EGFR, but very little is known on the resulting FSH- and EGF-dependent network. We built a single network using data underlying both networks. This leads to a new hypothesis on the activation of MEK by p38MAPK, which we validate experimentally. These preliminary results represent a first step in the demonstration of a cross-talk between these two major MAP kinases pathways.

Highlights

Within a multicellular organism, the behavior of individual cells is partly controlled by hormones that bind to their cognate receptors
A specific experimental result will lead to a specific interpretation, i.e. to an instantiated piece of biological knowledge, but the manner wherewith it is interpreted does not depend on this particular result
Our method tries to reproduce the reasoning made by biologists when interpreting experimental results, by automatically deducing new knowledge using a set of expert rules that make explicit this reasoning

Summary

Introduction

The behavior of individual cells is partly controlled by hormones that bind to their cognate receptors. Rather than depicting influences of molecules on each other (e.g. A activates B), these pathways are detailed representations of the molecular mechanisms that come into play (e.g. A phosphorylates B, and phosphorylated B is active) These pathways are represented using standards, such as the Systems Biology Graphical Notation (SBGN) Process Description language[5] (PD), which helps their understanding[6]. A third category, relying on the use of experimental data, adopts a radically different point of view, by using reasoning techniques to interpret qualitative data[20,21] Those methods define expert rules that allow interpreting experimental data in order to build gene regulatory networks[20] or executable models of signaling pathways[21]. Another current limitation of this method is the lack of rules to infer relationships from perturbation experiments (such as those involving inhibitors, siRNA, knock-outs or mutants), which are widely used in signaling biology

Methods

Results

Conclusion