ShinyButchR: Interactive NMF-based decomposition workflow of genome-scale datasets

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Non-negative Matrix Factorization (NMF) is valuable for genomic data analysis, enabling feature extraction and signature identification.
• Traditional NMF analysis presents challenges including complex tool installation, a steep learning curve, and significant computational time.

Purpose of the Study:

• To develop an accessible R/Shiny application for comprehensive NMF-based genomic data analysis.
• To simplify matrix decomposition, feature extraction, visualization, and signature identification using NMF.

Main Methods:

• Development of ShinyButchR, an R/Shiny application integrating NMF analysis.
• Utilized the R package ButchR, featuring TensorFlow solvers for NMF algorithms.
• Incorporated methods for determining optimal factorization rank and a novel feature selection strategy.

Main Results:

• ShinyButchR provides a complete workflow for NMF-based genomic data analysis.
• The application facilitates interactive visualization and association of identified signatures with biological and clinical variables.
• The ButchR package offers enhanced NMF solvers and downstream analysis functions.

Conclusions:

• ShinyButchR significantly lowers the barrier to entry for NMF analysis in genomics.
• The application empowers researchers to perform advanced NMF analyses, including signature identification and association studies.
• ShinyButchR and ButchR enhance the utility and accessibility of NMF for genomic research.