ExTILAR – Network Inference


The Transcription factor binding site Integrating LARS (TILAR) algorithm is a network inference algorithm that uses a linear model to infer static gene regulatory networks based on the Least Angle RegreSsion (LARS) [1]. Using the TILAR concept of modeling, a regulating gene affects the expression of the regulated gene via a transcription factor (TF) only if the regulated gene has a transcription factor binding site (TFBS) for the TF, while the regulating gene does not. Thus, the TILAR algorithm makes use of a variety of prior-knowledge types such as TFBS information as well as text-mining knowledge about the relations between the genes and the TFs. Therefore, TILAR created networks are composed of two types of nodes - genes and TFs - and two types of edges - prior-knowledge TF-to-gene relations and gene-to-TF relations.

ExTILAR extends the TILAR algorithm by adding a variety of new aspects to the modeling concept to support time resolved data. In this tutorial we will show how to infer a dynamic transcription factor network using time series data of cultivated murine hepatocytes.

The data used was obtained from Zellmer et al., 2010 [2] and monitors the cellular response of murine hepatocytes to culture medium exchange at 5 time points (0, 3, 6, 12, 24 hours) with three replicates each. Details about pre-processing, clustering and filtering of the differentially expressed genes are outlined in Vlaic et al., 2012 [7]. The final data set is consisting of 22 differentially expressed genes for which the corresponding protein is known to exhibit transcription factor activity and 6 cluster profiles representing functional modules.


There are 4 objects stored in the RData file:

es: The expression set that contains all measurements and meta-data.
realtions: The TF-to-gene relations extracted from Transfac and PathwayStudio, or predicted by oPOSSUM.
knowledge: Known gene-to-TF interactions derived by text-mining using PathwayStudio.
genesI: A character vector that provides the mapping from the gene names used in the relations and knowledge tables to the used rownames of the measurements in the expression set.

The algorithm works on ExpressionSet objects of the Biobase package available from Bioconductor [3]. The key idea is to collect all relevant data about the experiment in the meta-data data-frame that can be obtained using the pData function. In that, columns define the experimental properties important for the ExTILAR algorithm while rows denote for the values of the properties for each microarray. i.e., the meta-data data-frame used here contains 6 important columns.

Replicate: Indicates the replicate number of the corresponding microarray.
Time: Indicates the time point the microarray corresponds to.
Input: Indicates the input function that should be used for the simulation.
exponentialIdentifyer: If the input function was chosen to be exponential, this value defines the the exponent e in the function exp^(e * time).
Interpolation: Indicates the interpolation method that should be used to obtain equidistant measurements.
InterpolationStep: Indicates the step-width for the interpolation.

In addition to the meta-data, ExTILAR needs a object of the type ParameterObject which combines parameter values used during the inference process. If no value for a certain parameter is given by the user, default values are automatically set. To create the transcription factor network presented in the publication [7] the following command has to be used:

o = extilar.create_ExTILAR_object(expressionSet=es, genes=genesI, relations=relations, knowledge=knowledge, parameterObject=p)

During this process, redundant relations are compressed and the knowledge is adapted. The final command starts the ExTILAR algorithm using the stepwise forward selection procedure for structure optimization:

result.fs = extilar.perform_fs(extilarObject=o)

This command automatically performs the network inference and stores the extracted networks in the folder the program was started in. The network can now be imported into Cytoscape [8] for further processing.

A general advantage of ODE-based models is the possibility of in silico knock-downs. To provide an example, we show how ExTILAR can be used to perform such knock-downs on single genes. Therefore, the corresponding gene is substituted by a linear decreasing input function modeling a continuous si-RNA knock-down. At first, the model has to be generated which can be done using the command:

= extilar.create_simulation_object(result=result.fs$res$res_pp, ci=result.fs$ci, experiment=1)

The sim Object contains now all information needed to numerically solve the set of ordinary differential equations. To knock-down Tgif1, the following command is used:

sim.kd = extilar.knock_down_gene(input=sim, gene="TGIF1", func=function(t){return(-(1/24)*t)})

Now, the set of differential equations can be numerically solved using the command:

fit.kd = extilar.simulate_network(input=sim.kd, t=1000, time=24, experiment=1)

The fit.kd object contains now the dynamics that can be plotted using the command:

extilar.plot_simulation(input=fit.kd, mat=result.fs$ci$mat[[1]], mat.mean=result.fs$ci$mat.mean[[1]], mat.sd=result.fs$ci$mat.sd[[1]], separatePlots=T, originalData=result.fs$ci$originalData[[1]])


This program is free software; you can redistribute it and/or modify it under the terms of theGNU General Public License as published by the Free Software Foundation, Version 3.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the impliedwarranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. ExTILAR.r contains source-code of from the R package lars_0.9-7 which is released under the GNU GPL v2.URL: http://www-stat.stanford.edu/~hastie/Papers/#LARS

I agree with the download the license:


[1] Hecker M, Goertsches RH, Engelmann R, Thiesen HJ, Guthke R: Integrative modeling of transcriptional regulation in response to antirheumatic therapy. BMC Bioinformatics 2009, 10:262, [http://dx.doi.org/10.1186/1471-2105-10-262].

[2] Zellmer S, Schmidt-Heck W, Godoy P, Weng H, Meyer C, Lehmann T, Sparna T, Schormann W, Hammad S, Kreutz C, Timmer J, von Weizsäcker F, Thürmann PA, Merfort I, Guthke R, Dooley S, Hengstler JG, Gebhardt R: Transcription factors ETF, E2F, and SP-1 are involved in cytokine-independent proliferation of murine hepatocytes. Hepatology 2010, 52(6):2127-2136, [http://dx.doi.org/10.1002/hep.23930]

[3] Bioconductor: Open software development for computational biology and bioinformatics R. Gentleman, V. J. Carey, D. M. Bates, B.Bolstad, M. Dettling, S. Dudoit, B. Ellis, L. Gautier, Y. Ge, and others 2004, Genome Biology, Vol. 5, R80

[4] Matys V, Fricke E, Geffers R, Gössling E, Haubrock M, Hehl R, Hornischer K, Karas D, Kel AE, Kel-Margoulis OV, Kloos DU, Land S, Lewicki-Potapov B, Michael H, Münch R, Reuter I, Rotert S, Saxel H, Scheer M, Thiele S, Wingender E: TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 2003, 31:374-378.

[5] Nikitin A, Egorov S, Daraselia N, Mazo I: Pathway studio-the analysis and navigation of molecular networks. Bioinformatics 2003, 19(16):2155–2157.

[6] Ho Sui SJ, Fulton DL, Arenillas DJ, Kwon AT, Wasserman WW: oPOSSUM: integrated tools for analysis of regulatory motif over-representation. Nucleic Acids Res 2007, 35(Web Server issue):W245-W252, [http://dx.doi.org/10.1093/nar/gkm427].

[7] Sebastian Vlaic, Wolfgang Schmidt-Heck, Madlen Matz-Soja, Eugenia Marbach, Jörg Linde , Anke Meyer-Baese , Sebastian Zellmer , Reinhard Guthke , Rolf Gebhardt : The Extended TILAR Approach: A novel Tool for Dynamic Modeling of the Transcription Factor Network Regulating the Adaption to in vitro Cultivation of Murine Hepatocytes . In submission

[8] Smoot, M. E.; Ono, K.; Ruscheinski, J.; Wang, P.-L. & Ideker, T.: Cytoscape 2.8: new features for data integration and network visualization. Bioinformatics, Department of Medicine, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093, USA. msmoot@ucsd.edu, 2011, 27, 431-432