KEDRI - GNetXP - Gene Network Explorer


AUT Main Nav
Centre Banner
Main Content

GNetXP - Gene Network Explorer

Team: Dr. Zeke Shun Heng Zhang, Vishal Jain and Prof. Nikola Kasabov

What is a Gene Regulatory Network?

A single gene interacts with many other genes in this process, inhibiting, directly or indirectly, the expression of some of them, and promoting others at the same time. This interaction can be represented as a gene regulatory network (GRN). GRN dynamically evolve and change their structure based on DNA and environmental information. The discovered gene regulatory networks (GRN) from time series of gene expression observations can be used in principal to (a) identify important genes in relation to a disease or a biological function, (b) gain an understanding on the dynamic interaction between genes, (c) predict gene expression values at future time points, (d) predict drug effect over time.

Software developed at KEDRI for GRN modelling: “HybridClust” GNetXP – “Gene Network Explorer

GNetXP – “Gene Network Explorer”.
Figure 1: GNetXP – “Gene Network Explorer”

GNetXP is a software system for gene time course data clustering and gene interaction network discovery. While developing this system our objective was to combine the strength of Genetic Algorithm and Expectation Maximization algorithm to produce a global yet efficient clustering algorithm.

At the higher level, GA searches for the optimal subset of genes that act as initial cluster centers and at the lower level, the local learning method, Expectation Maximization algorithm (EM), performs local clustering from these initial centers. These two learning algorithms are implemented with mixture of Multiple Linear Regression models (MLRs) (A) Standard Expectation Maximization (EM) algorithm that uses random initialized cluster centres and (B) Hybrid Genetic Algorithm (GA) and EM that uses GA for initializing the cluster centres.

Using GA, the hybrid algorithm searches the clustering solution space more thoroughly, offering more consistent and more optimal solutions than the standard EM algorithm by far. Our software extracts GRN using two stage process (1) Hybrid Genetic Algorithm and Expectation Maximization algorithm is applied on clustering the large number of gene trajectories using the mixture of multiple linear regression models for fitting the trajectory data (2) Kalman Filter (parameter estimation) is applied to identify a set of first-order differential equations that describe the dynamics of the representative trajectories, and use these equations for discovering important gene interactions and predicting gene expression values at future time points.

Demonstration results from the clustering and Kalman filter modules are shown in figure 2 and 3 respectively.

Clustering with GNetXP.
Figure 2: Clustering with GNetXP

Predictions using Kalman filter algorithm.
Figure 3: Predictions using Kalman filter algorithm

Using HybridClust on Your Own Data

To use HybridClust, one must save the time series gene expression file into an appropriate structure (figure 4). The time points must be expressed as a row vector and stacked on top of the gene expression data. The matrix form of the data structure thus appears as follows.

Using HybridClust on Your Own Data.
Figure 4: Using HybridClust on Your Own Data

This matrix should then be saved as an ascii file. HybridClust will accept most ascii file format, including delimited and fixed width partitioning.


GNetXP is freely available for academic users and is for “non-commercial” use only. You can download this software system here.

Kindly note that the software system provided as MATLAB PCODE. You will need a copy of MATLAB to run this software.


For more information on background on GRN and methodology implemented in GNetXP software system, please refer:

2004 - N. Kasabov, Z. S. H. Chan, V. Jain, I. Sidorov, and D. S. Dimitrov, "Gene Regulatory Network Discovery from Time-Series Gene Expression Data - A Computational Intelligence Approach," Lecture Notes in Computer Science, vol. 3316/2004, pp. 1344-1353.

2005 - N. Kasabov, Z. S. H. Chan, V. Jain , I. Sidorov and D. S. Dimitrov, Computational Modelling of Gene Regulatory Networks, Book Chapter in "Series on Advances in Bioinformatics and Computational Biology (Information Processing and Living Systems)" Eds. Vladimir B Bajic, Tan Tin Wee, Vol. 2, pp 673-686.

2005 - Z. Chan, N. Kasabov and L. Collins, A hybrid genetic algorithm and expectation maximization method for global gene trajectory clustering, Journal of Bioinformatics and Computational Biology, Imperial College Press, Vol. 3, No. 5 (2005) 1227-1242.

2004 - Zeke S. H. Chan and Nikola Kasabov, Gene Trajectory Clustering with a Hybrid Genetic Algorithm and Expectation Maximization Method in International Joint Conference on Neural Networks, IJCNN 2004, Budapest, 16-30 June 2004, IEEE Press

2004 - Z. Chan, N. Kasabov, and L. Collins, "A two-stage methodology for gene regulatory network extraction from time-course gene expression data," presented at IEEE Workshop on Biomedical Applications of Circuits and Systems, Singapore, 2004.

Last updated: 29 Feb 2012 12:30pm

Auckland University of Technology, New Zealand | Copyright © | Privacy | Site map