tinge

Name

TINGe -- Tool Inferring Networks of GEnes

Synopsis

tinge-mi {-i file} options...

Description

TINGe is a parallel tool for constructing gene regulatory networks from large-scale gene expression data. It uses information theoretic criteria and statistical testing to detect dependencies between genes, and it can be run on large parallel machines, such as IBM Blue Gene. TINGe is implemented in C++ and MPI, therefore it should be executed in the MPI environment by using an adequate mpiexec command.

Options

In general, TINGe tries to maintain compatibility of options, and file formats, with ARACNe. Please refer to [1] for the detailed description of file formats.

-i file

Read expression profiles from file.

-o file

Write output network to file. If option is not specified, the output file is created by changing extension of the input file to .adj.

-j file

Read relevance network from file.

-w file

Store relevance network in file.

-l file

Read list of transcription factors from file.

-a estimator

Set the estimator that should be used to calculate mutual information to estimator. Currently two estimators are provided: B for B-spline estimator [2], and G for Gaussian kernel estimator [3]. We highly recommend using the B-spline estimator for its high performance and accuracy. By default the B-spline estimator is used.

-b bins

Set the number of bins used by the B-Spline mutual information estimator to bins. This option is ignored if Gaussian kernel estimator is used. By default 10 bins are used. See [2] for a more detailed explanation.

-k order

Set the order of basis B-Spline functions used by the mutual information estimator to order. This option is ignored if Gaussian kernel estimator is used. Default is 4. See [2] for a more detailed explanation.

-p pval

Set the significance level of testing if mutual information is statistically greater than 0 to pval. This option is mutually exclusive with -t. Default is 1.0, that is, no test is performed. See [4] for a more detailed explanation.

-t threshold

Set the threshold value for mutual information to threshold. This option is mutually exclusive with -p.

-e tolerance

Set the tolerance of data processing inequality to tolerance. Default is 1.0, that is, no processing is applied. See [1] for a more detailed explanation.

-C method

Convert final mutual information using method. C converts mutual information to correlation coefficient using formula by Joe [5]. Currently only this method is supported. By default no conversion is applied.

-r size

Set the number of generated bootstrap networks to size. This option is mutually exclusive with -y. Default is 0, that is, no bootstrapping is applied.

-y size

Set the number of columns used for analysis to size. If set, only size randomly selected columns (with no replacement) are used. This option is mutually exclusive with -r. Default is 0, that is, no sampling is applied.

-x seed

Set random seed utilized in statistical tests to seed. By default seed is generated from the current time and process identifier.

-m

Turn on memory usage reports. By default reports are disabled. On some architectures memory reports might be incorrect.

-v

Show progress, i.e. notify about iteration progress during the main processing phase.

-h

Show summary of options.

Example

	mpiexec -np 32 tinge-mi -i tost.exp -p 0.001 -e 0.1
      

The above command will execute tinge-mi on 32 processors. A network will be created based on the expression data from tost.exp and it will be stored in tost.adj. Statistical testing at significance level of 0.001 will be performed, and data processing inequality will be applied with tolerance of 0.1.

References

[1] Margolin, A.A. et al., "Reverse engineering cellular networks", Nature Protocols, vol. 1 no. 2, 2006.

[2] Daub, C.O. et al., "Estimating mutual information using B-spline functions - an improved similarity measure for analysing gene expression data", BMC Bioinformatics, vol. 5 no. 118, 2004.

[3] Moon, Y., Rajagopalan, B., Lall, U., "Estimation of mutual information using kernel density estimators", Physical Review E., vol. 52 no. 3, 1995.

[4] Zola, J., Aluru, M., Sarje, A., Aluru, S., "Parallel information theory based construction of genome wide gene regulatory networks", In IEEE Transactions on Parallel and Distributed Systems, vol. 21, no. 12, pp. 1721-1733, 2010.

[5] Joe, H., "Relative entropy measures of multivariate dependence", Journal of the American Statistical Association, vol. 84 no. 405, 1989.

Bugs

Bugs? What bugs? Well, if you are sure that you have found a bug you can contact Jaroslaw Zola.

Author

Jaroslaw Zola <jaroslaw.zola@gmail.com>, Maneesha Aluru, Dan Nettleton, Srinivas Aluru.

Copyright

TINGe (C) 2007-2012 Jaroslaw Zola under GNU GPLv3 License. The build system, (c) 2006-2010 Jaroslaw Zola under the MIT License. The JAZ code, (c) 2004-2010 Jaroslaw Zola under the Boost Software License. The MPIX code, (c) 2005-2008 Jaroslaw Zola under the Boost Software License.