satijalab/seurat v4.0.0

Version 4.0.0

on GitHub

Changes

• Removed old, deprecated code
• Utilize SeuratObject as base of object code
• Default neighbor finding algorithm changed from "rann" to "annoy"
• Default ncells parameter in SCTransform() changed to 5000
• Default fold change in FindMarkers() changed from ln to log2
• Implementation improvements to AverageExpression()
• AnchorSet class re-implemented as a virtual class from which IntegrationAnchorSet and TransferAnchorSet now inherit.
• Point size in VlnPlot() now set automatically if not specified
• Return the sample.tree properly when integrating with a single reference dataset
• Replace as.character.quosure usage with as_label due to deprecation
• Minor changes to the exact calculation of the anchor weight matrix
• Default rasterization limit in DimPlot() and FeaturePlot() changed from 50,000 to 100,000
• SCTransform() now returns a formalized Assay subclass SCTAssay()
• When using normalization.method='SCT' in FindTransferAnchors(), normalize query using reference SCT model when possible.

Additions

• Expose FoldChange() component in FindMarkers()
• Add IntegrateEmbeddings() to correct embeddings of DimReducs
• Add ProjectUMAP() to project query cells into a reference UMAP space
• Add MapQuery() as a wrapper around IntegrateData(), IntegrateEmbeddings(), and ProjectUMAP()
• Add MappingScore to compute a per-cell mapping score used in Azimuth
• Add AggregateExpression() for summation based pseudobulk calculations
• Add mixscape functionality via CalcPerturbSig(), PrepLDA(), RunLDA(), DEenrichRPlot(), MixscapeHeatmap(), MixscapeLDA(), PlotPerturbScore(), RunMixscape()
• Add FindSubCluster() to further cluster existing clusters
• Add supervised PCA functionality via RunSPCA()
• Add functionality to enable weighted nearest neighbor analyses via FindMultiModalNeighbors()
• Add neighbor visualization plot via NNPlot().
• Add PredictAssay() to impute expression or embeddings from nearest neighbors
• Add AnnotateAnchors() to aid in AnchorSet interpretation as well as subset.AnchorSet()
• Add flexibility of choice for cell column in Read10X()
• Add rasterization option to FeatureScatter() and VariableFeaturePlot()