Biological processes in every living organism are governed by complex interactions between thousands of genes and gene products. Modeling and understanding these interactions is essential to progress in many important research areas such as medicine, agriculture or more recently synthetic biology. Such interactions are typically modeled as gene regulatory networks, whose inference is one of the principal challenges in systems biology. Rapid advances in high-throughput biotechnologies, for instance microarrays, enable simultaneous measurement of the expression levels of all genes in a given organism. The goal of gene network reverse engineering is to reconstruct the network from observations of gene expression data. The task is challenging as there are tens of thousands of genes in any complex organism, and in many cases it is impossible to find a reliable way to limit analysis to only a subset of them.
TINGe – Tool for Inferring Networks of GEnes, is a parallel and multi-platform framework for reconstructing gene regulatory networks from large gene expression data. It uses parallel processing, information theoretic criteria and statistical testing to derive networks with thousands of genes from microarray sets with thousands of observations. TINGe has been used to reconstruct a whole-genome network of Arabidopsis thaliana from 3,546 microarray measurements, which comprises of 15,495 genes. TINGe is (or has been) developed by Jaroslaw Zola, Maneesha Aluru and Srinivas Aluru with contributions from Abhinav Sarje and Dan Nettleton.
Network inference is just a first step in gaining insights into regulatory mechanisms of a given organism. Taking into account that any complex organism contains thousands of genes it becomes challenging to isolate sub-networks that would specifically characterize selected biological processes. GeNA – GEne Networks Analyzer, is a tool that for a given set of “seed” genes uses gene ranking mechanism to extract subnetworks of genes with similar biological function. GeNA uses algorithm akin to the topic-sensitive PageRank and it has been implemented as a stand-alone tool and as a plugin for Cytoscape. GeNA is developed by Jaroslaw Zola.
Jam Make Redux – the build system used by TINGe depends on Jam which is found in majority of Linux and Unix distributions. If not it can be obtained here.
C++ compiler – we tested majority of C++ compilers, including GNU, Intel, Oracle Solaris Studio Express, IBM and PGI. Any standard conforming compiler should work equally well.
MPI library – TINGe is a parallel software. Even if you plan to run it on a single CPU/core you will still need MPI-2 compliant library to compile it. We tested MPICH2, OpenMPI, MVAPICH2, Oracle Message Passing Toolkit (formerly Sun HPC ClusterTools).
Cytoscape – GeNA is a Cytoscape plugin. To execute it you will need Cytoscape 2.6.3 or newer.
The latest version of TINGe is 1.061. TINGe is distributed under GNU General Public License version 3. Mpiext and jaz components are distributed under Boost Software License version 1.0. The x-build build system is distributed under the MIT License.
The bundle above provides the source code that can be compiled on any common architecture (PC, PC cluster, IBM BlueGene, etc.) or it can be used to derive specialized implementations. It contains a comprehensive documentation explaining how to install and use TINGe.
The current version of GeNA is 0.1. It is provided as a Cytoscape plugin. GeNA is distributed under GNU General Public Lincese version 2.
Which paper I should cite when using TINGe?
Can I run TINGe on the Cell processor?
Yes! TINGe is available for Cell processors in version 1.015. This version lacks certain functionality offered by the newer TINGe releases but it can be run on clusters of Cells. See this paper for more details. To obtain the Cell enabled source code please contact Jaroslaw Zola or Srinivas Aluru.
What are the file formats used by TINGe?
How many observations/microarrays I need to run TINGe?
TINGe will run with as few as two observations. However, mutual information estimators used by TINGe require that many observations are provided. Usually, we recommend more than 100.
How long will it take to run TINGe with my data?
If you have n genes/probes and m observations/microarrays, and you use p processors, then you can expect
that the execution time will be bounded by O(n^2 * m / p). To get an estimate for your hardware, extract a small number of genes from your dataset and run it with TINGe. Next, use the measured time to estimate the runtime for the entire dataset. Note also that TINGe provides on the fly estimate when executed, and
additionally it may show progress information (enabled by -v option).
I am getting “Error: data is too small” message, why?
TINGe requires that your input expression data has at least two rows and two columns. If your data is bigger
than that, which we hope is the case, this error indicates that the input file is malformed. Most likely because you are using spaces and not tabs to separate columns.
Which paper I should cite when using GeNA?
Please cite our paper in Nucleic Acids Research.
I am getting a strange message when starting GeNA in Cytoscape, what should I do?
It means that the binary version of the GeNA engine that is packaged with the plugin cannot be run
on your architecture. You will have to compile GeNa from sources, or contact us to obtain the updated
version that will be compatible with your system.