ordinalbayes: Fitting Ordinal Bayesian Regression Models to High-Dimensional Data Using R.

Area of Science:

• Genomics
• Biostatistics
• Cancer Research

Background:

• Cancer staging is crucial for treatment decisions.
• Identifying molecular features associated with cancer stage can reveal tumor aggressiveness and therapeutic targets.
• High-throughput genomic data, like RNA-Seq from The Cancer Genome Atlas Project (TCGA), offers potential for such discoveries.

Purpose of the Study:

• To introduce the ordinalbayes R package for fitting cumulative logit models to high-dimensional ordinal data.
• To enable variable selection in situations where the number of predictors exceeds the sample size (P > N).
• To apply the package to The Cancer Genome Atlas cervical cancer (TCGA-CESC) dataset to identify molecular features associated with cancer stage.

Main Methods:

• Development and application of the ordinalbayes R package, built upon the runjags R package.
• Utilizing penalized Bayesian ordinal response models for variable selection.
• Fitting cumulative logit models to high-throughput RNA-Seq and clinical data from TCGA-CESC.

Main Results:

• The ordinalbayes package effectively fits models to high-dimensional datasets.
• The package successfully performs variable selection, identifying molecular features associated with cervical cancer stage.
• Demonstrated utility of the package on the TCGA-CESC dataset.

Conclusions:

• The ordinalbayes R package provides a valuable tool for analyzing high-dimensional genomic data with ordinal outcomes.
• It facilitates the identification of molecular features linked to cancer aggressiveness and potential therapeutic targets.
• The package enhances the analysis of complex biological datasets, particularly in cancer research.