dimensionality_reduction: Dimensionality reduction suite

In vittoriofortino84/COPS: Clustering algorithms for Omics-driven Patient Stratification

Dimensionality reduction suite

Description

Applies multiple dimensionality reduction techniques to input data.

Usage

dimensionality_reduction(
  x,
  dimred_methods = c("pca", "umap"),
  output_dimensions = NULL,
  pca_dims = c(2),
  umap_dims = c(2),
  tsne_perplexities = c(45),
  tsne_pca = TRUE,
  umap_neighbors = 20,
  initial_dims = 50,
  ...
)

Arguments

x	Data matrix, features on columns and samples on rows.
dimred_methods	Vector of method names, see details for options.
output_dimensions	Vector of dimensionalities to compute using each applicable method.
pca_dims	PCA specific output dimensions.
umap_dims	UMAP specific output dimensions.
tsne_perplexities	Vector of t-SNE perplexity settings to generate embeddings with.
tsne_pca	Whether to apply PCA before t-SNE, which massively boosts performance.
umap_neighbors	UMAP parameter, affects manifold computation.
initial_dims	Number of principal components used in t-SNE and UMAP.
...	Extra arguments are ignored.

Details

Method options

• "none" - Returns original data as is

• "pca" - Principal Component Analysis

• "tsne" - t-distributed stochastic neighbor embedding

• "umap" - Uniform Manifold Approximation and Projection for Dimension Reduction

Value

Returns a list of embeddings, elements are named based on methods used

vittoriofortino84/COPS documentation built on Jan. 28, 2025, 3:16 p.m.