GeneNetwork
, written in C++, offers a series of modules to allow experimentalists
to explore and visualize gene network inferred through the analysis
of data generated from microarray experiments. The architecture and
method of GeneNetwork are outlined in Figure 1. Overall, GeneNetwork
works as follows:
- GeneNetwork
receives experimental data in the tab-delimited text format
(Figure 2)
- proceed
data interpolation through interpolation controller while there
is not enough data points to start the inference
- implement
reverse engineering inference approaches through modeling controller
to discover causal relationships and produce the regulation
matrix between genes
- automatically
visualize the network based on the regulation matrix
- compare
the inferred intuitive network with on-line database (such as
KEGG and Biocarta), relied on the information from network graphical
viewer and information viewer
- the
user may review the proposed sets of experiments and generate
biological hypotheses
Fig 1. The architecture of GeneNetwork
Fig 2. Input data format
Data Interpolation
The
gene network inference techniques will cover one thing in common:
the size of the data points required increases while we increase
the number of variables to mathematical model. To increase the amount
of data from which to start network inference, we can start the
network analysis with proceeding interpolation of the gene expression
time series data points. The interpolation controller (Figure 3)
of GeneNetwork provides three kinds of data interpolation approaches:
Linear Interpolation, Lagrange Polynomial Interpolation, and Cubic
Spline Interpolation.
Fig 3. Interpolation controller
Modeling
Controller (Network structure Inference)
GeneNetwork
supports four genetic network inference models to extract the "gene
regulation matrix" from the gene expression data, which cover
the three categories: 1).Linear model; 2).S-system; 3).Boolean network;
4).dynamic Bayesian network. In the software, genetic algorithm
is applied to conquer the convergence to local optimal on searching
nonlinear solution space and learning network structure. A user
can manipulate the four reverse engineering approaches through themodeling
controller (Figure 4) of GeneNetwork to infer the gene regulation
matrix, which describes which genes regulate each other and how
the genetic inputs affect gene expression, from a massive set of
gene expression data. In the supplemental document, we provide a
detailed description the four approaches and an application on the
Saccharomyces cerevisiae cell cycle gene expression data.
Fig 4. Modeling controller
Network
Graph Viewer / Information Viewer
After
generating a gene regulation matrix, there are still some problems
to be solved in how to extract valuable information from the matrix.
For this reason, GeneNetwork embraces several default network visual
layouts and several graphical setting options. A network diagram
is represented as a graph with nodes corresponding to genes and
edges indicating relations between gene network components. Nodes
are represented by colourful circles, and interaction are represented
by colourful lines and arrows. Information on the network structure
and genes, which taken from the gene regulation matrix and input
information, can be shown on information viewer. Clicking on any
node reveals the biological processes that involve selected gene
and its relation to other genes. GeneNetwork is also fully customizable
and can allow a user to define personal setting for generation of
interaction networks by manipulating several graphical setting options,
such as: linkage change, gene selection, gene search, font and graph
setting etc.
Fig
5. Representative network layout
|
