vrnmf: Volume-Regularized Structured Matrix Factorization
Implements a set of routines to perform structured matrix factorization with minimum volume constraints. The NMF procedure decomposes a matrix X into a product C * D. Given conditions such that the matrix C is non-negative and has sufficiently spread columns, then volume minimization of a matrix D delivers a correct and unique, up to a scale and permutation, solution (C, D). This package provides both an implementation of volume-regularized NMF and "anchor-free" NMF, whereby the standard NMF problem is reformulated in the covariance domain. This algorithm was applied in Vladimir B. Seplyarskiy Ruslan A. Soldatov, et al. "Population sequencing data reveal a compendium of mutational processes in the human germ line". Science, 12 Aug 2021. <doi:10.1126/science.aba7408>. This package interacts with data available through the 'simulatedNMF' package, which is available in a 'drat' repository. To access this data package, see the instructions at <https://github.com/kharchenkolab/vrnmf>. The size of the 'simulatedNMF' package is approximately 8 MB.
||R (≥ 3.5.1)
||graphics, ica (≥ 1.0), lpSolveAPI (≥ 126.96.36.199), Matrix, nnls, parallel (≥ 3.5.1), quadprog (≥ 1.5), stats
||knitr (≥ 1.28), rmarkdown (≥ 2.1), testthat
||Ruslan Soldatov [aut],
Peter Kharchenko [aut],
Viktor Petukhov [aut],
Evan Biederstedt [cre, aut]
||Evan Biederstedt <evan.biederstedt at gmail.com>
Please use the canonical form
to link to this page.