Unify Dimensionality Reduction Results

Dimensionality reduction is widely used in many domains for analyzing and visualizing high-dimensional data. 'tidydr' provides uniform output and is compatible with multiple methods, including 'prcomp', 'cmdscale', 'Rtsne', 'umap' and 'metaMDS'. Any function returning a numeric matrix can also be used. The unified result can be visualized directly with 'ggplot2', and several methods can be run and compared in a single call.

tidydr

tidydr provides uniform output and is compatible with multiple methods for dimensionality reduction, including prcomp, cmdscale, Rtsne, umap and metaMDS. Any function that returns a numeric matrix can also be used.

Installation

Install the released version from CRAN:

install.packages("tidydr")

Install the development version from GitHub:

remotes::install_github("YuLab-SMU/tidydr")

Usage

library(ggplot2)
library(tidydr)

x <- dr(data = iris[, 1:4], fun = prcomp)
autoplot(x, aes(color = Species), metadata = iris[, 5, drop = FALSE]) +
    theme_dr()

The methods known to work are listed by available_methods(). Methods that need their own arguments receive them through ..., e.g. dr(iris[, 1:4], Rtsne::Rtsne, check_duplicates = FALSE).

Comparing several methods

dr_compare() runs several methods on the same data. Each method is evaluated independently, so a method that fails is reported in the summary instead of aborting the whole call:

r <- dr_compare(iris[, 1:4],
                funs = list(prcomp = stats::prcomp, umap = uwot::umap),
                dim = 1:2)

r$summary      # method, status, n, k, has_eigenvalue, has_stress, error
autoplot(r)    # one facet per method that produced coordinates

Clustering and silhouette widths

nk() computes the average silhouette width over one or more values of k. It uses cluster::pam() by default; any other clustering function can be plugged in through fun:

si <- nk(iris[, 1:4], 2:4)                     # pam(), the default
si <- nk(iris[, 1:4], 3, fun = stats::kmeans)  # any other clusterer

autoplot(si)                                   # average silhouette width vs k
autoplot(si, k = 3)                            # samples coloured by cluster
autoplot(si, k = 3, type = "silhouette")       # per-cluster silhouette bars

The per-sample widths of a single k are available from silinfo_widths():

w <- silinfo_widths(si, 3)
head(w)

Going further

For more examples, see vignette("tidydr"), the GitHub repository, and the CRAN page.