Open AccessShinyButchR: Interactive NMF-based decomposition workflow of genome-scale datasets
Author(s) -
Andrés Quintero,
Daniel Hübschmann,
Nils Kurzawa,
Sebastian Steinhauser,
Philipp Rentzsch,
Stephen Krämer,
Carolin Andresen,
Jeongbin Park,
Roland Eils,
Matthias Schlesner,
Carl Herrmann
Publication year - 2020
Publication title -
biology methods and protocols
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.555
H-Index - 5
ISSN - 2396-8923
DOI - 10.1093/biomethods/bpaa022
Subject(s) - non negative matrix factorization , interpretability , computer science , matrix decomposition , workflow , visualization , feature extraction , feature selection , data mining , identification (biology) , artificial intelligence , rank (graph theory) , feature (linguistics) , signature (topology) , pattern recognition (psychology) , machine learning , mathematics , database , linguistics , eigenvalues and eigenvectors , physics , botany , philosophy , quantum mechanics , combinatorics , biology , geometry
Non-negative matrix factorization (NMF) has been widely used for the analysis of genomic data to perform feature extraction and signature identification due to the interpretability of the decomposed signatures. However, running a basic NMF analysis requires the installation of multiple tools and dependencies, along with a steep learning curve and computing time. To mitigate such obstacles, we developed ShinyButchR, a novel R/Shiny application that provides a complete NMF-based analysis workflow, allowing the user to perform matrix decomposition using NMF, feature extraction, interactive visualization, relevant signature identification, and association to biological and clinical variables. ShinyButchR builds upon the also novel R package ButchR, which provides new TensorFlow solvers for algorithms of the NMF family, functions for downstream analysis, a rational method to determine the optimal factorization rank and a novel feature selection strategy.