RunMDS: Run MDS

In nbroguiere/burgertools: Collection of tools extending the functionality of Seurat

Run MDS

Description

Compute a multidimensional scaling (MDS) dimensionality reduction on a Seurat object, using the provided distance metric. By default, using 30 dimensions in the reduced space, based on the variable features of the current default assay and data layer, with results stored under the name "mds".

Usage

RunMDS(
  object,
  distance = NA,
  n.dims = 30,
  features.use = NA,
  assay = NA,
  layer = "data",
  reduction.name = "mds",
  key = "MDS_",
  batch_size = 400,
  n.cores = 1,
  seed = 42
)

Arguments

object	Seurat object.
distance	Distance function, e.g. FractionDifferingVariants for variants or stats::dist() for common distances applying to continuous data. NA will use FractionDifferingVariants if the assay used is "VAR", euclidean dist otherwise.
n.dims	numeric(1). The number of dimensions in the reduction (Default:30).
features.use	character(n). The features to use to compute the reduction. NA for most variable features of assay (Default: NA).
assay	character(1). The assay to use. NA uses "VAR" if available, otherwise the DefaultAssay (Default:NA).
layer	character(1). The Seurat object assay layer to use (Default:"data").
reduction.name	character(1). The name under which the dimensionality reduction will be stored under (Default: "mds").
key	character(1). The key (prefix to column numbers) used to name the MDS coordinates (Default: "MDS_")
batch_size	numeric(1). The number of cells processed for each batch of the divide and conquer MDS. The batches are then realigned and merged. Smaller batches provide faster runs, larger batches provide more accurate results (Default: 400).
n.cores	numeric(1). The number of cores to use for parallel processing (Default: 1). Unix only.
seed	numeric(1). Random seed set at the beginning of the run for reproducible results (Default:42).

Details

Wrapper of the bigmds divide and conquer method, see bigmds::divide_conquer_mds() for details and references.

For usage on variant data in vartrix-like sparse matrix conventions, use the custom distance function FractionDifferingVariants().

For use on continuous rather than discrete data, typically consider stats::dist().

Value

Returns the Seurat object.

Examples

DefaultAssay(MySeuratObject) <- "VAR" # variant calls, sparse matrix with values 1,2,3 for ref/alt/heterozygous in vartrix conventions. Assuming VariableFeatures have been set to most informative variants for this assay. Will also use VAR by default if assay=NA, so no need to explicitely change the default assay.
MySeuratObject <- RunMDS(MySeuratObject)
DimPlot(MySeuratObject, reduction="mds")
MySeuratObject <- RunUMAP(MySeuratObject,reduction="mds",dims=1:30)
DimPlot(MySeuratObject, reduction="umap")

nbroguiere/burgertools documentation built on Jan. 30, 2024, 3:48 a.m.