A novel algorithm for finding optimal driver nodes to target control complex networks and its applications for drug targets identification

Wei-Feng Guo,Luonan Chen,Qian-Qian Shi,Tao Zeng,Cheng-Ming Zhang,Shao-Wu Zhang

doi:10.1186/s12864-017-4332-z

Abstract

BackgroundThe advances in target control of complex networks not only can offer new insights into the general control dynamics of complex systems, but also be useful for the practical application in systems biology, such as discovering new therapeutic targets for disease intervention. In many cases, e.g. drug target identification in biological networks, we usually require a target control on a subset of nodes (i.e., disease-associated genes) with minimum cost, and we further expect that more driver nodes consistent with a certain well-selected network nodes (i.e., prior-known drug-target genes).ResultsTherefore, motivated by this fact, we pose and address a new and practical problem called as target control problem with objectives-guided optimization (TCO): how could we control the interested variables (or targets) of a system with the optional driver nodes by minimizing the total quantity of drivers and meantime maximizing the quantity of constrained nodes among those drivers. Here, we design an efficient algorithm (TCOA) to find the optional driver nodes for controlling targets in complex networks. We apply our TCOA to several real-world networks, and the results support that our TCOA can identify more precise driver nodes than the existing control-fucus approaches. Furthermore, we have applied TCOA to two bimolecular expert-curate networks. Source code for our TCOA is freely available from http://sysbio.sibcb.ac.cn/cb/chenlab/software.htm or https://github.com/WilfongGuo/guoweifeng.ConclusionsIn the previous theoretical research for the full control, there exists an observation and conclusion that the driver nodes tend to be low-degree nodes. However, for target control the biological networks, we find interestingly that the driver nodes tend to be high-degree nodes, which is more consistent with the biological experimental observations. Furthermore, our results supply the novel insights into how we can efficiently target control a complex system, and especially many evidences on the practical strategic utility of TCOA to incorporate prior drug information into potential drug-target forecasts. Thus applicably, our method paves a novel and efficient way to identify the drug targets for leading the phenotype transitions of underlying biological networks.

Highlights

The advances in target control of complex networks can offer new insights into the general control dynamics of complex systems, and be useful for the practical application in systems biology, such as discovering new therapeutic targets for disease intervention
An interesting question, known as target control problem of complex networks, is posed that how can we chose the driver variables from the system to control a subset of the whole nodes about a trim point [8]
To evaluate the target control efficiency on an arbitrary network, we first introduce two factors αand β, which represent the ratio of the target nodes and constrained nodes to the whole network nodes respectively; To target control the target nodes O(α), our TCOA can identify the optional nodes set D(α, β) with the minimum driver nodes and the maximum quantity of driver nodes contained in a given constrained set Q(β).‖D(α, β)‖/‖O(α)‖ denotes the ratio of the quantity of identified drivers to the quatity of targets, and ‖D(α, β) ∩ Q‖/‖D(α, β)‖ denotes the ratio of the quantity of identified drivers among constrained nodes to the quantity of all the identified drivers

Summary

Introduction

The advances in target control of complex networks can offer new insights into the general control dynamics of complex systems, and be useful for the practical application in systems biology, such as discovering new therapeutic targets for disease intervention. E.g. drug target identification in biological networks, we usually require a target control on a subset of nodes (i.e., disease-associated genes) with minimum cost, and we further expect that more driver nodes consistent with a certain well-selected network nodes (i.e., prior-known drug-target genes)

Objectives

Methods

Results

Conclusion