ordination {MatrixQCvis}R Documentation

Dimensionality reduction with ordination methods PCA, PCoA, NMDS, UMAP and tSNE

Description

The function 'ordination' creates a 'data.frame' with the coordinates of the projected data. The function allows for the following projections: Principal Component Analysis (PCA), Principal Coordinates Analysis/Multidimensional Scaling (PCoA), Non-metric Multidimensional scaling (NMDS), t-distributed stochastic neighbor embedding (tSNE), and Uniform Manifold Approimation and Projection (UMAP).

Usage

ordination(x, type = c("PCA", "PCoA", "NMDS", "tSNE", "UMAP"), params = list())

Arguments

x

'matrix', containing no missing values, samples are in columns and features are in rows

type

'character', specifying the type/method to use for dimensionality reduction. One of 'PCA', 'PCoA', 'NMDS', 'tSNE', or 'UMAP'.

params

'list', arguments/parameters given to the functions 'stats::prcomp', 'stats::dist', 'Rtsne::Rtsne', 'umap::umap'

Details

The function 'ordination' is a wrapper around the following functions 'stats::prcomp' (PCA), 'stats::cmdscale' (PCoA), 'vegan::metaMDS' (NMDS), 'Rtsne::Rtsne' (tSNE), and 'umap::umap' (UMAP). For the function 'umap::umap' the method is set to 'naive'.

Value

'tbl'

Author(s)

Thomas Naake

Examples

x <- matrix(rnorm(1:10000), ncol = 100)
rownames(x) <- paste("feature", 1:nrow(x))
colnames(x) <- paste("sample", 1:ncol(x))
params <- list(method = "euclidean", ## dist
    initial_dims = 10, max_iter = 100, dims = 3, perplexity = 3, ## tSNE
    min_dist = 0.1, n_neighbors = 15, spread = 1) ## UMAP
ordination(x, type = "PCA", params = params)
ordination(x, type = "PCoA", params = params)
ordination(x, type = "NMDS", params = params)
ordination(x, type = "tSNE", params = params)
ordination(x, type = "UMAP", params = params)


[Package MatrixQCvis version 1.1.2 Index]