In insilico/Rinbix: Insilico Bioinformatics Toolbox

The Rinbix package is a collection of functions that support the In silico research group's bioinformatics toolbox for network and machine learning analysis in bioinformatics (inbix). These functions are grouped roughly into the following categories:

• Feature Selection and Ranking
• Classification and Prediction
• Genetic Association Interaction Networks (GAIN)
• Clustering and Network Community Detection
• Data Simulation
• Interfacing with the C++ inbix Program for Performance
• RNA-Seq Classification and Prediction Pipeline

These functions are documented in the package vignettes and reference manual.

Installation

Rinbix is a package that is most easily installed through RStudio's Tools->Install Packages... menu. Alternatively, the package can be installed from an R console with:

install.packages("path_to_file/Rinbix-VERSIONSTRING.tar.gz", 
                 repos = NULL, type = "source")

or from the command line with:

R CMD INSTALL Rinbix-VERSIONSTRING.tar.gz

Conventions

Variable and function names are in camelCase. Data frames for case-control data sets are referred to as 'regressionData', 'trainData' and 'testData' and contain subject rows and variables (gene) columns, The last column is special and contains the case-control phenotype, coded 0/1. Due to the bioinformatics application domain and common data mining and machine learning jargon, the following terms are used interchangeably:

• rows of X, subjects, samples, obervations
• columns of X, variables. predictors, attributes, SNPs, genes
• last column, y, response, phenotype

Quick Start Example

This simple example demonstrates a workflow from data simulation to ranked genes. Simulate a differential co-expression data set with 10 genes and 40 subjects.

Data simulation

library(Rinbix)
data("scaleFreeNetwork")
datasetObj <- createDiffCoexpMatrixNoME(M = 10, N = 40, meanExpression = 7, 
                                        A = Rinbix::scaleFreeNetwork,
                                        randSdNoise = 0.5, sdNoise = 0.05,
                                        sampleIndicesInteraction = seq(1:3))
dataset <- as.data.frame(datasetObj$regressionData)
colnames(dataset) <- colnames(datasetObj$regressionData)
rownames(dataset) <- rownames(datasetObj$regressionData)
cat("Simulated genes:", colnames(dataset)[1:3], "\n")

First few rows of the simulated data set

knitr::kable(dataset)

Run a ranker/feature selector algorithm. dcGAIN + SNPrank

dcgain_snprank_results <- rankDcgainSnprank(dataset)

dcGAIN + SNPrank results

knitr::kable(dcgain_snprank_results)

Run a ranker/feature selector algorithm. reGAIN + SNPrank

regain_snprank_results <- rankRegainSnprank(dataset)

reGAIN + SNPrank results

knitr::kable(regain_snprank_results)

insilico/Rinbix documentation built on June 7, 2020, 6:19 a.m.