VisuNet is an interactive tool for structural analysis of complex rule-based classifiers. VisuNet can be applied to any classification problem and is commonly used with complex health-related decision tasks. The rule networks produced can clearly identify driving genes (metabolites, methylation sites, etc) and their expression levels. VisuNet is implemented in R and uses the Shiny Gadgets attributes. The tool includes construction, filtration, visualization and customization of networks from rule-based models. VisuNet is available on GitHub.

Rule network construction

A rule network is constructed from sets of IF-THEN rules. In the network, nodes are conjuncts of rules, ie. features, and edges connect nodes with corresponding conjuncts in rules.

The rule network resulting from a rule

Installation

devtools::install_github("komorowskilab/VisuNet")

Example

This example uses gene expression data for young males with autism and control(Alter et al. (2011)). The rule-based classifier was created using R.ROSETTA (see Garbulowski et al. 2019).

require(VisuNet)

#Sample rule set for a classifier of young males with autism and control
#'Line by line' data type
autcon_ruleset

#Run VisuNet
#Remember to click DONE once you finish working on VisuNet
vis_out <- visunet(autcon_ruleset, type = 'L')

The sample rule network for the young males with autism and control classifier from VisuNet (constructed for min decision coverage=27% and min accuracy=88%)

Network legend

The rule network legend

References

Alter, Mark D., Rutwik Kharkar, Keri E. Ramsey, David W. Craig, Raun D. Melmed, Theresa A. Grebe, R. Curtis Bay, et al. 2011. “Autism and Increased Paternal Age Related Changes in Global Levels of Gene Expression Regulation.” Journal Article. PloS One 6 (2): e16715–e16715. https://doi.org/10.1371/journal.pone.0016715.

Garbulowski, Mateusz, Klev Diamanti, Karolina Smolińska, Patricia Stoll, Susanne Bornelöv, Aleksander Øhrn, and Jan Komorowski. 2019. “R.ROSETTA: A Package for Analysis of Rule-Based Classification Models.” Journal Article. bioRxiv, 625905. https://doi.org/10.1101/625905.

 

© 2019 Komorowski's BioInformatics Lab,Uppsala University Contact: Karolina Smolinska

Design by Yan Holtz