Changelog • parafac4microbiome

parafac4microbiome (development version)

CRAN release: 2025-07-31

Addition of age, gender and oral health data to vanderPloeg2024.
Addition of TIFN2, MAINHEALTH, GOHTRANS, and AP articles to package documentation. These are only rendered for the website due to computational requirements.

Removal of the all-at-once optimization for parafac() since the ALS scheme outperforms it under all circumstances.
Removal of deprecated verbose option for parafac().
Renaming of vignettes to be easier to visually parse (also in preparation of the addition of new vignettes).
Overhaul of the entire testing suite to be more robust and fast.

CRAN release: 2025-05-20

The package is now fully compatible with R version 4.5.
Due to breaking changes to the TreeSummarizedExperiment package that came with R version 4.5, minimum version requirements for TreeSummarizedExperiment, MicrobiotaProcess, and SummarizedExperiment have temporarily been added to the dependencies. This breaks compatibility with older versions of Ubuntu temporarily.

Added the function reshapeData() which reshapes a long dataframe of counts into a data cube.
Added tests for reshapeData().
Fujita2023, Shao2019, and vanderPloeg2024 now utilize reshapeData when they are being generated.
Many bugfixes across all functions in the package.

CRAN release: 2025-03-21

vanderPloeg2024 now contains all datasets from the original paper.
The contents in /data-raw/ are updated to better organize the studies.
The Frobenius norm of the tensor is now collapsed into the subject mode (was the O-norm).
Fixed an issue where flipLoadings() would not flip any of the modes if there was evidence that all three modes should be flipped.
Fixed an issue where flipLoadings() would not work when one model was provided.
All vignettes are updated to better reflect the updated functions for this version.

CRAN release: 2024-09-24

Minor changes to tests for parafac_gradient to make them more robust towards various platforms checked by CRAN.

CRAN release: 2024-09-17

Further rework of importPhyloseq, importTreeSummarizedExperiment and importMicrobiotaProcess to meet CRAN requirements.

Streamlined examples in importPhyloseq, importTreeSummarizedExperiment and importMicrobiotaProcess.

Edited DESCRIPTION to meet CRAN requirements.
importPhyloseq was modified to use rTensor-based cube folding instead of a for-loop.
importTreeSummarizedExperiment was modified to use rTensor-based cube folding instead of a for-loop.
importMicrobiotaProcess was modified to use rTensor-based cube folding instead of a for-loop.

‘parafac’ is now capable of running an all-at-once optimization using the methods=“opt” parameter. For now, the default remains methods=“als” (i.e. the ALS algorithm) because it converges faster to a similar solution.
importPhyloseq allows the user to import a phyloseq object for parafac modelling.
importTreeSummarizedExperiment allows the user to import a TreeSummarizedExperiment object for parafac modelling.
importMicrobiotaProcess allows the user to import a MicrobiotaProcess object for parafac modelling.
The text in the vignettes were updated to better reflect the changes per version 0.1.0.
The readme and vignettes figures now use sign flipping to make comparison with the paper easier.
Some documentation and testing changes anticipating a CRAN release.

parafac is now a custom function based on an ALS algorithm allowing for much more output (see documentation).
initializePARAFAC initializes the input vectors either randomly on based on a best-guess SVD model of the unfolded array.
parafac_core_als contains this ALS algorithm.
parafac_fun calculates the loss of a parafac model in anticipation of an all-at-once optimization implementation.
assessNumComponents and checkModelStability have been renamed into assessModelQuality and assessModelStability respectively to clarify their use.
checkModelStability now works with a minimum and maximum number of components.
checkModelStability reports Factor Match Score in a plot.
calculateFMS calculates pairwise Factor Match Scores for a list of model objects.
parafac, multiwayCenter, multiwayScale and multiwayCLR are now based on the new rTensor dependency for tensor unfolding
plotOverallTCCs has been merged with plotModelTCCs.
Many bugfixes. PARAFAC model solutions should be more stable as a result.
Many documentation changes across the board. Overall, the use case per function should be much clearer.
Vignettes are updated to reflect and utilize the new changes.
Removed dependency: paramGUI