Run PaCMAP (Pairwise Controlled Manifold Approximation)

Run PaCMAP (Pairwise Controlled Manifold Approximation)

Usage

RunPaCMAP(object, ...)

# S3 method for class 'Seurat'
RunPaCMAP(
  object,
  reduction = "pca",
  dims = NULL,
  features = NULL,
  assay = NULL,
  layer = "data",
  n_components = 2,
  n.neighbors = NULL,
  MN_ratio = 0.5,
  FP_ratio = 2,
  distance_method = "euclidean",
  lr = 1,
  num_iters = 450L,
  apply_pca = TRUE,
  init = "random",
  reduction.name = "pacmap",
  reduction.key = "PaCMAP_",
  verbose = TRUE,
  seed.use = 11L,
  ...
)

# Default S3 method
RunPaCMAP(
  object,
  assay = NULL,
  n_components = 2,
  n.neighbors = NULL,
  MN_ratio = 0.5,
  FP_ratio = 2,
  distance_method = "euclidean",
  lr = 1,
  num_iters = 450L,
  apply_pca = TRUE,
  init = "random",
  reduction.key = "PaCMAP_",
  verbose = TRUE,
  seed.use = 11L,
  ...
)

Arguments

object

An object. This can be a Seurat object or a matrix-like object.

...

Additional arguments to be passed to pacmap.PaCMAP.

reduction

The reduction to be used. Default is "pca".

dims

The dimensions to be used. Default is NULL.

features

The features to be used. Default is NULL.

assay

The assay to be used. Default is NULL.

layer

The layer to be used. Default is "data".

n_components

The number of PaCMAP components. Default is 2.

n.neighbors

A number of neighbors considered in the k-Nearest Neighbor graph. Default to 10 for dataset whose sample size is smaller than 10000. For large dataset whose sample size (n) is larger than 10000, the default value is: 10 + 15 * (log10(n) - 4).

MN_ratio

The ratio of the ratio of the number of mid-near pairs to the number of neighbors. Default is 0.5.

FP_ratio

The ratio of the ratio of the number of further pairs to the number of neighbors. Default is 2.

distance_method

The distance metric to be used. Default is "euclidean".

lr

The learning rate of the AdaGrad optimizer. Default is 1.

num_iters

The number of iterations for PaCMAP optimization. Default is 450.

apply_pca

Whether pacmap should apply PCA to the data before constructing the k-Nearest Neighbor graph. Using PCA to preprocess the data can largely accelerate the DR process without losing too much accuracy. Notice that this option does not affect the initialization of the optimization process. Default is TRUE.

init

The initialization of the lower dimensional embedding. One of "pca" or "random". Default is "random".

reduction.name

The name of the reduction to be stored in the Seurat object. Default is "pacmap".

reduction.key

The prefix for the column names of the PaCMAP embeddings. Default is "PaCMAP_".

verbose

Whether to print the message. Default is TRUE.

seed.use

The random seed to be used. Default is 11.

Examples

pancreas_sub <- Seurat::FindVariableFeatures(pancreas_sub)
#> Finding variable features for layer counts
pancreas_sub <- RunPaCMAP(
  object = pancreas_sub,
  features = SeuratObject::VariableFeatures(pancreas_sub)
)
#> Warning: Layer ‘data’ is empty
#> Error in h(simpleError(msg, call)): error in evaluating the argument 'x' in selecting a method for function 't': subscript out of bounds
CellDimPlot(
  pancreas_sub,
  group.by = "CellType",
  reduction = "pacmap"
)
#> Warning: No shared levels found between `names(values)` of the manual scale and the
#> data's fill values.