Movatterモバイル変換

rliger Next

Ability to reorganize datasets
- Allow doing something likereorganize(ligerObj, variable = "somethingNotDataset") andresulting in a new liger object with different ligerDatasetgrouping.
Ability to do downstream analysis on H5 data
- Pseudo-bulk should be easy because we are just aggregatingcells.
- Wilcoxon might be a bit harder because ranks are calculated per genebut the H5 sparse data is column majored. Might need to find a faston-disk transposition method.

Added naive GSEA analysis on factor gene loading (W) to test if anyknown gene sets (e.g. cell cycle) is enriched in any factor. ImplementedinfactorGSEA().
Added dense data loading support for H5AD files
Optimized obs metadata parsing for H5AD files
Fixed ggplot2 color picking when coloring by logical value
Fixed H5AD file layer detecting bug
Fixed some other minor bugs

Implemented highly efficient on-disk iNMF that scales to a millioncells using slightly more time than in-memory version, requiring onlylaptop-level memory.
Added 10X H5 data and H5AD loading function that loads the data intoregular dgCMatrix in memory or the DelayedArray representation backed ondisk, the latter is used for on-disk iNMF implementation.
AddedselectBatchHVG() which implements another HVGselection strategy, credit to SCIB
AddingsuggestK() back with new methodology
Clarified optimalrunGOEnrich() workflow and added foldenrichment metric in the returned result
Fixed important bug in online iNMF scenario 2
Fixed multiple problems related to ATAC analysis
- Fixed Wilcoxon rank-sum test bug when using ATAC peak counts
- Fixed gene coordinate parsing bug from BED file
- Optimized peak parsing speed

AddedcentroidAlign() for new cell factor loadingalignment method
AddedplotProportionBox() for visualizing compositionalanalysis
AddedplotClusterGeneViolin() for visualizing geneexpression in clusters
AddedplotBarcodeRank() for basic QC visualization
AddedplotPairwiseDEGHeatmap() for visualizing pairwiseDEG results
AddedplotGODot() for visualizing GO enrichmentresults
AddedcalcNMI() for evaluating clustering resultsagainst ground truth
AddedligerToH5AD() allowing reticulate/Python freeexport of liger object to H5AD format. This is presented in extensionsource code (i.e. not loaded withlibrary(rliger)).
Added organism support inrunGeneralQC() and refinedhemoglobin gene matching regex pattern.
Optimized DE test memory usage scalability for both pseudo-bulkmethod and wilcoxon test
OptimizedplotProportionPie() by adding argumentcircleColors
OptimizedplotVolcano() text annotation positioning andgene highlighting logic.
Optimized visualization function additional argumentdocumentation
ChangedrunMarkerDEG() andrunPairwiseDEG() default method from"wilcoxon" to"pseudoBulk"
FixedrunMarkerDEG(method = "pseudobulk") bug inassigning pseudo-replicates, and optimized error/warning signaling.
Fixed bug incalcAlignment(),subsetMemLigerDataset(),cellMeta()
Fixed bug in old version updating functions

Fixed wrong UINMF aborting criteria
Fixed example/test skipping criteria for non-existingdependencies
Fixed file access issue when checking on CRAN
Updated installed data filesystem.file("extdata/ctrl.h5", "extdata/stim.h5") to be ofstandard 10X H5 format
UpdatedquantileNorm() automatic reference selectionaccording to #297
Other minor fixes (including #308)

AddedligerDataset class for per-dataset informationstorage, with inheritance for specific modalities
Added a number of plotting functions with clear function names anduseful functionality
Added Leiden clustering method, now as default rather thanLouvain
Added pseudo-bulk DEG method
Added DEG analysis with one-vs-rest marker detection inrunMarkerDEG() and pairwise comparison inrunPairwiseDEG()
Added gene name pattern for expression percentage QC metric
Added native Seurat object support for the core integrationworkflow
Added a documentation website built with pkgdown
Added new iNMF variant method, consensus iNMF (c-iNMF), inrunCINMF(). Not stable.
Added GO enrichment dowsntream analysis inrunGOEnrich()
Changedliger object class structure
Moved iNMF (previouslyoptimizeALS()), UINMF(previouslyoptimizeALS(unshared = TRUE)) and online iNMF(previouslyonline_iNMF()) implementation to new packageRcppPlanc with vastly improved performance. Now wrapped inrunINMF(),runUINMF() andrunOnlineINMF() respectively, and all can be invoked withrunIntegration().
Updated H5AD support to match up with Python anndata package 0.8.0specs
Renamed many function/argument names to follow camelCase style,original names are still available while deprecation warnings areissued