The geneNetBP Homepage

The geneNetBP package provides functions to fit Conditional Gaussian Bayesian network (CG-BN) and Discrete Bayesian network to genotype-phenotype or Quantitative Trait Loci (QTL) data, absorb evidence in the genotype-phenotype network and quantify and visualize the changes in network beliefs. The package facilitates the smooth transition from QTL mapping to graphical modeling to system-wide prediction after absorbing network evidence. Specifically, geneNetBP can be used to absorb and propagate phenotypic evidence through a given CG-BN representation of a genotype-phenotype network, compute the updated beliefs across the network, quantify the effects (Jeffrey’s Signed Information & Fold Changes) and provide visualizations for interpretation.

Download

Download the latest version 2.0.1 of geneNetBP from CRAN.

The geneNetBP package version 1.0.0 is archived and will no longer be supported. It is however available for download on CRAN.

Installation

For CG-BN implementation, the geneNetBP package makes extensive use of RHugin package that provides an R interface for the Hugin Decision Engine (HDE) - a commercial software developed by HUGIN EXPERT A/S for building and making inference from Bayesian belief networks. The RHugin package provides an interface to communicate with the HDE from within R.

For learning and belief propagation in CG-BN, the geneNetBP package requires HDE and the R package RHugin to be installed.

Installation Steps

1. Install Hugin Decision Engine:

The geneNetBP package is compatible with the free demo version of Hugin Researcher/Developer, Hugin Lite Demo that can be obtained from Download Free HUGIN. You can also access this page by visiting Hugin Lite homepage, and navigating to Products -> Services -> Training. The ftp site for accessing older versions: here . Note that the free demo version is limited to handle only 50 states and 500 cases.

If you do not have Hugin installed in the default location you will need to set the HUGINHOME environment variable before using the RHugin package. Also, you will need to modify the HUGINHOME variable. Please see the instructions on the project homepage for RHugin .

2. Install RHugin package:

RHugin is available on R-Forge and NOT on CRAN. Installation instructions for RHugin can be found on its project homepage RHugin .It is important to install the matching versions and the architecture (32/64 bit) of Hugin Lite, RHugin and R as listed on RHugin project homepage. Please note that RHugin is required for proper functioning of CG-BN implementation geneNetBP . The package RHugin will not automatically load upon loading geneNetBP . Please use library(RHugin) or require(RHugin) to load it before using geneNetBP.

3. Install other package dependencies:

Discrete bayesian networks learning and inference is implemented using algorithms from the packages bnlearn and gRain. geneNetBP depends on these packages which are available on CRAN. These packages should get installed automatically with geneNetBP. You can manually install them by using R install command

 install.packages("bnlearn")

 install.packages("gRain")

In addition to these, both geneNetBP and RHugin depend on the Bioconductor packages graph and Rgraphviz. Run the following commands to install them.

  	source("http://bioconductor.org/biocLite.R")
  	biocLite(c("graph", "Rgraphviz"))

Note that geneNetBP v2.0.0 and higher do not depend on the package scales unlike v1.0.0.

4. Install the geneNetBP package:

The geneNetBP package latest version 2.0.1 can be downloaded from CRAN.

Documentation

The methods for predicting and visualizing system-wide effects of genotype-phenotype networks under perturbations are described in our publication Moharil J. et.al. (2016)

The package vignette illustrates the methods with examples (including how to reproduce results from the publication) and provides tutorials on each function. [geneNetBPv2.0.0-vignette]

A complete list of currently available functions and datasets included in the package. [geneNetBPv2.0.0-manual]

The belief propagation (stable scheme) is described in Lauritzen S.L and Jensen F.

References

: Belief Propagation in genotype-phenotype networks (2016); Moharil, J., May, P.,Gaile, D.P. and Blair, R.H.; Stat Appl Genet Mol Biol. , Vol. 15(1), pp. 39-53
: Stable local computation with conditional gaussian distributions (2001); Lauritzen, S. L. and Jensen F.; Stat. Comput., Vol. 11(2), pp. 191-203
: Propagation of probabilities, means, and variances in mixed graphical association models (1992); Lauritzen, S. L.; J. Am. Statist. Assoc., Vol. 87, pp. 1098-1108
: Local computations with probabilities on graphical structures and their application to expert systems (1988); Lauritzen, S. L. and D. J. Spiegelhalter; J. Roy. Stat. Soc. B Met., Vol. 50 (2), pp. 157-224