This repository provides a simple example pipeline to perform Bayesian probabilistic network inference using the RIMBANet software.
Usage:
-
Download folders
bin
,script
, andPerls
. Place them in a directory say$HOME/user/RIMBANet/
. Add eXecute permission to files in the two folders. Lastly, either add$HOME/user/RIMBANet/bin
to your systemPATH
variable or movebin/testBN
to a place included inPATH
. -
Prepare discretized data matrix (discretized continuous values into ordinal
0
,1
, and2
; see an example using kmeans clustering based discretization method) and save in a tab-delimited text file nameddata.discretized.txt
in your working directory, say/path/to/working/directory
. In this file, each row represents a variable, and each column represents a sample except the first columns which consists of variable names. However, there should be no header row. -
Call RIMBANet to construct networks with different random starting structures. A wraper script
caller.bsh
is provided to create all required files from the data and then submit parallel jobs on a LSF environment. To change any of the default behabivors, please editcaller.bsh
,BNJob.bsh
, andBNJobbase.bsh
. To add additional prior information, please refer to Perl scripts in folderPerls
.
bash $HOME/user/RIMBANet/script/caller.bsh /path/to/working/directory
- When LSF jobs (default
1000
) are completed successfully, create a consensus network structure (default final outputresult.links3.links.txt
) consists of de-looped links present in, by default, at least 30% of the 1000 individual structures.
bash $HOME/user/RIMBANet/script/createStructure.bsh bn.param.txt