Bugfix to rownames of mnnCorrect() output when using correct.all=TRUE with subset.row=.
Migrate findMutualNN() to BiocNeighbors.
Support d=NA in multiBatchPCA() for more convenient disabling of PCA in calling functions.
Bugfix for d=NA with specified subset.row= in fastMNN().
Added the applyMultiSCE() function to easily apply functions across main/alternative Experiments from multiple SingleCellExperiment inputs.
Added the mnnDeltaVariance() function to compute diagnostics from the variances of the differences between MNN pairs.
Added the quickCorrect() function to quickly perform intersection, normalization, feature selection and correction.
Added some clustering-based diagnostics (clusterAbundanceVar(), clusterAbundanceTest() and compareMergedClusters()) from the OSCA book.
File-backed matrices are now realized into memory prior to multiBatchPCA().
Allow regressBatches() to operate without batch= when design= is provided.
Added d= and related options to conveniently perform a PCA on the ResidualMatrix.
Added correct.all= option to all correction functions for consistency.
Added a deferred=TRUE default to multiBatchPCA and its callers,
to encourage use of deferred matrix multiplication for speed.
Switched default PCA algorithm in multiBatchPCA to IrlbaParam.
Added add.single= mode for endomorphic addition of correction results in correctExperiments().
Support the use of arbitrary design matrices in regressBatches().
Allow lists of objects to be directly passed into the ... for many functions.
Added the clusterMNN() function for performing MNN on cluster centroids.
Added get.variance= option to fastMNN() to return variance explained from PCA.
Support d=NA to skip the PCA step altogether.
Modified correctExperiments() to preserve non-conflicting rowData() fields.
Deprecated rotate.all= in favour of get.all.genes= in multiBatchPCA().
Switched BSPARAM= to use IrlbaParam(deferred=TRUE) by default in fastMNN(), so that the default behaviour is actually fast.
Deprecated auto.order= in favor of merge.order= and auto.merge= in fastMNN() and mnnCorrect(). Automatic merging now detects potential tree-based merges. Merge trees can also be specified as input.
Added the correctExperiments() function to cbind the original assays alongside the merged values.
Added the subset.row= argument to cosineNorm() for in-place subsetting.
Added batch= and preserve.single= arguments to multiBatchNorm(). Standardized behavior of subset.row= by adding a normalize.all= argument.
Added the regressBatches() function for correction via standard linear regression.
Added the prop.k= argument in all MNN-related functions, to allow the value of k to adapt asymmetrically to the size of each batch.
New package batchelor, for batch correction of single-cell (RNA sequencing) data.