A Multiscale Analysis of Cell Polarity Transitions in a Bacterium (2): using higher-order biological networks to explore temporal patterns in cellular oscillators

Host laboratory and collaborators

Bianca Habermann / IBDM, UMR 7288, Marseille / bianca.HABERMANN@univ-amu.fr

Laurent Tichit / I2M, UMR 7373, Marseille / laurent.tichit@univ-amu.fr

Alain Barrat / CPT, UMR 7332, Marseille / alain.barrat@gmail.com


Cell decisions are governed by complex regulatory networks that integrate changes in the environment and convey a cellular response. Understanding the functioning of these networks is a major challenge because they are generally obtained from fragmentary datasets that lack both quantitative and spatiotemporal information. As a result, the genetic pathways mostly consist of blueprints that capture the interactions between the components but generally fail to contain mechanistic and thus predictive value. Given that molecular interactions are in essence only amenable to low-throughput analyses, any first attempt in modeling dynamic networks must focus on highly tractable experimental systems. In this project, five CENTURI teams, Mignot and Michelot (proposal 1), and Habermann, Tichit and Barrat (proposal 2) will develop an interdisciplinary collaboration to elucidate how bacterial cells (Myxococcus xanthus) make directional decision using an evolutionarily-conserved G-protein cell polarity oscillator.

These two intertwined projects will combine biophysical cell mimetic assays, genetic live experiments and computational simulations of higher-order networks to provide a first spatiotemporal model of the protein-protein interactions that drive protein oscillations. In the long run, this work will serve as a framework to study single cell decisions in multicellular contexts, a question of general significance in higher organisms.

This second proposal is principally computational.


G-protein, biochemical oscillator, motility, cell mimetic systems, higher order networks


The correct and timely interaction of molecules is essential for all cellular functions. In Myxococcus xanthus, as an example, cells change the direction of their movement by sequential and spatial interactions of three key molecules, MglA, MglB and RomR, which are regulated by a receptor-activated signal transduction pathway, the Frz pathway. The interactions of proteins within a cell are usually represented as a first order protein-protein interaction network (PPIN), where proteins are the nodes that are connected if they interact. Yet, the representation of PPIs as simple networks has limitations. It does not consider PPIs that are mutually exclusive and have to follow a specific timely order. In this project, we plan to use higher-order networks to represent mutually exclusive binding in protein-protein interaction networks (PPINs). We will compare biological higher-order networks (HON) to classical binary PPI networks and specifically test whether the information gain of HONs can address the problem of temporal order in cellular PPIs. We will use available knowledge on the regulatory system controlling the Frz pathway, as well as data on temporal and spatial order of binding between key components emanating from proposal 1 to build a HON as a first spatio-temporal model of protein-protein interaction driving protein oscillation. We will examine its predictive power and test resulting hypotheses experimentally.

Proposed approach (experimental / theoretical / computational)

We will follow the subsequent steps to build a HON of the Frz pathway: 1) Reconstruct the Frz pathway, including known binding events outside the pathway. 2) Identify spatio-temporal binding events, fed by data from proposal 1. 3) Construction of the HON using the known concurrent and consecutive interactions. Starting from an ordinary network (ON) with nodes i and links (ij) representing interactions between nodes, HONs are built with nodes (ij) representing the links in the ON, and links (ij->jk) representing causally admissible successions of links (interactions) of the ON. 4) Comparison of the performance of the HON to classical PPINs and multi-layered networks; this includes a) the adaption of community-design algorithms for HONs; b) comparison of results from the 3 network types w.r.t. current biological knowledge (benchmark: M. xhantus extended Frz-pathway); c) comparison of 3 network types w.r.t. hub node bias, or anti-correlated expression of nodes. Resulting predictions will be passed on to proposal 1 for experimental testing.


The project is highly interdisciplinary as it links in vitro biophysics to genetics and high-end computational modeling. Specifically, two postdocs for 2 years will be recruited and shared between five CENTURI member teams (Michelot, Mignot, Habermann, Tichit, Barrat). The postdoc selected for the proposal 1 will perform most of the experimental work, and will work side-by-side with the members of the proposal 2 to implement the computational algorithm.

All expertise for the project is present in the three partners. The postdoc of proposal 2 will be supervised by research team leaders located at a biological (IBDM), a mathematical (I2M) and a physics institute (CPT). He/she will furthermore work in close collaboration with the experimental postdoc from proposal 1.

The techniques employed range from classical bioinformatics (structure- and sequence-based identification of mutually exclusive binding partners, literature curation), to graph theory and mathematical modeling (construction of HONs), to HON flux and network analysis.

A successful candidate will have a major in physics or computational science with a strong motivation for addressing biological questions.