CIF: Small: Novel biologically inspired methods for analyzing multilayer networks

Complex systems are often organized as multiple networks of interactions, where different characteristics such as interaction mechanisms or external perturbations govern the topology of each network. Civil infrastructures, social networks, producer/consumer/retailer networks are just a few examples to these systems. Each network in such a system shows how entities interact with each other under certain premises and conditions, while the interactions across different networks show how different conditions affect the entire system.

The investigator calls such complex systems multilayer networks. To understand how these systems work, it is of utmost importance to consider the entire system of networks holistically, rather than each network at different layers independently as they collectively describe the functionality of the underlying system. The main objective of this project is to develop the fundamental tools needed to study multilayer networks, which requires creation of novel computational techniques for motif identification and comparative analysis of such networks.

Biological networks, such as cellular networks, are naturally organized in multiple layers, where each layer describes the interaction topology among a set of molecules under a unique set of restrictions, such as external or internal stress condition, cell type, developmental stage, and interaction type. Thus, using biological systems as a network model provides opportunities to describe and study complex network systems.

Computational analysis of interactions among different molecules organized as a biological network is a computationally interesting and difficult problem. Studying collections of such networks through multilayer networks introduces further challenges as (i) the interaction topologies as well as the interaction types among molecules may vary across different layers, and (ii) networks at different layers of a multilayer network may interact as they may share some molecules or the molecules at different layers may affect each other.

Following from these observations, the PI is tackling the below two goals to achieve the main objective. (1) Identify building blocks in multilayer networks. (2) Develop tools for comparative analysis of multilayer networks. Both motif identification and comparative network analysis problems for classical single layer networks have been considered in the literature for over a decade.

There are however very limited studies which address these challenges for complex multilayer systems. The methods developed in this project are combining and extending the theory and the algorithms for fundamental graph theory, computational biology, bioinformatics, and machine learning to achieve these objectives.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.