tsne: tsne: A t-SNE function

Description Usage Arguments Value Examples

View source: R/tsne.R

Description

This is a flexible t-SNE function that can be run on a standard data frame. It is a wrapper for Rtsne/ggplot2 code and can be customised with different colours and font sizes and more.

Usage

1
2
3
4
5
6
7
tsne(mydata, labels = FALSE, perplex = 15, printres = FALSE,
  seed = FALSE, axistextsize = 18, legendtextsize = 18,
  dotsize = 5, textlabelsize = 4, legendtitle = "Group",
  controlscale = FALSE, scale = 1, low = "grey", high = "red",
  colvec = c("skyblue", "gold", "violet", "darkorchid", "slateblue",
  "forestgreen", "violetred", "orange", "midnightblue", "grey31", "black"),
  printheight = 20, printwidth = 22, text = FALSE)

Arguments

mydata

Data frame or matrix: if dataframe/matrix should have samples as columns and rows as features

labels

Character vector: if we want to just label with gender for example

perplex

Numerical value: perplexity value that Rtsne uses internally

printres

Logical flag: whether to print the t-SNE into current directory

seed

Numerical value: optionally set the seed

axistextsize

Numerical value: axis text size

legendtextsize

Numerical value: legend text size

dotsize

Numerical value: dot size

textlabelsize

Numerical value: text inside plot label size

legendtitle

Character vector: text legend title

controlscale

Logical flag: whether to control the colour scale

scale

Numerical value: 1=spectral palette, 2=manual low and high palette, 3=categorical labels

low

Character vector: continuous scale low colour

high

Character vector: continuous scale high colour

colvec

Character vector: a series of colours in vector for categorical labels, e.g. c("sky blue", "gold")

printheight

Numerical value: png height

printwidth

Numerical value: png width

text

Character vector: if we wanted to label the samples with text IDs to look for outliers

Value

A t-SNE plot object

Examples

1
TSNE <- tsne(mydata,perplex=15)

crj32/M3C documentation built on Feb. 19, 2020, 11:39 p.m.