Changes in muscomic objects to optimize storage and computation, by avoiding long-format tables in favors of sparse matrices and by reworking the object structure. muscadet and muscomic objects from previous versions are not supported by this version.
CreateMuscomicObject() now expects input mat_counts as cells x features (instead of features x cells).
muscomic object new structure:
coverage slot now contains counts and logratio layers, each with:
mat: matrix of counts or log ratio cells × features (sparse or dense depending on data type).
coord.features: coordinates of features matching the matrix.
label.features: feature type labels matching counts or log ratio.
allelic slot now contains 3 elements:
mat.RD: sparse matrix of reference allele counts cells x features.
mat.AD: sparse matrix of alternate allele counts cells x features.
coord.vars: coordinates of variant positions matching the matrices. Implemented new function makeAllellicSparse() to transform of long-format table (input) into sparse matrices. Updated addAlleleCounts()CreateMuscomicObject() for sparse format.
All package functions have been updated consequently to support the new format.
Implemented aggregateCounts() to replace mergeCounts() and support the new format.
Updated function examples, tests, and documentation consequently.
clusterMuscadet() with Seurat-based clustering now supports leiden_method argument:
Default "igraph" implementation; "leidenbase" also available.
Seurat dependency set to v5.3.1.
Added random.seed argument for reproducible clustering (default set to 1).
Updated example dataset objects exdata: subset of cells and subset of features/variant on 3 chromosomes for object size optimization. Updated function examples and vignettes consequently.
Fix, move (from Imports to Suggests) and remove some dependencies.
muscadet 0.1.3
Bug Fixes
Fixed issue on methods applied to one-omic muscadet objects to get back to methods returning a list of objects instead of returning the object itself in the one-element list case.
Fixed the table of features coordinates for ATAC in computeLogRatioATAC(), now the windows not covered by peaks are dropped.
Fixed issue about nX number of chromosomes in cnaCalling() to take the maximum number of chromosomes depending on genome and not on the number of chromosomes in the object (case with missing chromosomes).
muscadet 0.1.2
Bug Fixes
Seurat dependency set to 5.3.0 at minimum to avoid the installation of leidenalg python package when selecting the Leiden method for clustering.
The clusters and ncells elements of the @cnacalling slot in muscadet objects are now updated during the cnaCalling() function right after some clusters need to be skipped due to not enough data. This resolves discordance between clusters and ncells elements with other @cnacalling elements, and avoids issues in plotting functions.
The CNA calling functions (assignClusters(), mergeCounts() and cnaCalling() now all in cnaCalling.R) now include a reset of the @cnacalling slot of the muscadet object, to avoid discordance between results if clusters or parameters are changed in between the different steps of the CNA calling.
Documentation
Documentation added to clusterMuscadet() and cluster_seurat() for Leiden clustering algorithm recommendation when using the Seurat method.
muscadet 0.1.1
Bug Fixes
Fix issue of heatmapMuscadet() crashing corner case: heatmapMuscadet() no longer crashes when averages=TRUE for one cluster only (monoclonal samples).
Fix issue of heatmapMuscadet() with a wrong cluster order after Seurat clustering: cluster_seurat() and imputeClusters() now return sorted clusters vectors, from 1 to n, cells with imputed clusters are therefore placed at the end of each cluster group.
Fix issue of wrong silhouette computation (due to clusters sorting changes): clusters are ordered according to the distance matrix order of cells used for silhouette computation, giving the correct silhouette scores.
The SeuratObject package is now attached along with muscadetto easily use Cells() and Features() methods on muscadet objects.
