Package 'scUtils'

Title: Utility Functions for Single-Cell RNA Sequencing Data
Description: Analysis of single-cell RNA sequencing data can be simple and clear with the right utility functions. This package collects such functions, aiming to fulfill the following criteria: code clarity over performance (i.e. plain R code instead of C code), most important analysis steps over completeness (analysis 'by hand', not automated integration etc.), emphasis on quantitative visualization (intensity-coded color scale, etc.).
Authors: Felix Frauhammer [aut, cre], Simon Anders [ctb] (Simon Anders wrote the colVars_spm function.)
Maintainer: Felix Frauhammer <[email protected]>
License: GPL-3
Version: 0.1.0.9000
Built: 2025-02-02 03:40:27 UTC
Source: https://github.com/felixthestudent/scutils

Help Index

Closed breaks for log scale

Description

Finds breaks that are powers of 2, and forces inclusion of upper and lower limits (displaying the closed interval). Including limits specifically is particularly useful for ggplot2's color/fill, as it emphasizes the meaning of maximal/minimal color intensities (see examples).

Usage

closed_breaks_log2(lims)

Arguments

lims

Vector with lower and upper limits (in that order) of the data that you want breaks for.

Details

The feat function uses closed_breaks_log2 to color by gene expression, where the maximal expression gives valuable intuition for a gene's overall expression strength. For x- or y-axis (scale_*_log10), I still recommend breaks_log from the scales package.

Value

Numeric vector with breaks.

See Also

closed_labels

Examples

# closed breaks include maximum, breaks_log do not:
closed_breaks_log2(lims = c(.01, 977.1))
scales::breaks_log()(c(.01, 977.1))

Human-readable labels for closed breaks

Description

Complements the closed_breaks_log2 function.

Usage

closed_labels(x, min_is_zero = FALSE)

Arguments

x

Vector of breaks for which to produce labels. Typically, this is the output of closed_breaks_log2.
min_is_zero

Should the smallest break be displayed as zero (TRUE) or as the actual value (FALSE). Default: FALSE

Details

This is a helper for the feat function. feat replaces numeric zeros with the next-smallest expression value to avoid taking the logarithm of zero. min_is_zero can be used to display the lowest break of the color scale as zero in these cases.

Value

Character vector with labels, used by feat function.

See Also

label_scientific label_number_auto

Examples

# human readable output:
 closed_labels(c(.001111,.122, 0.5, 10, 100, 1800))

Variance computation for sparse matrices

Description

Compute variance for each column / each row of a dgCMatrix (from Matrix package).

Usage

colVars_spm(spm)

rowVars_spm(spm)

Arguments

spm

A sparse matrix of class dgCMatrix from the Matrix package.

Details

The only supported format currently is dgCMatrix. While the Matrix package has other formats, this one is used for scRNAseq raw count data. Function code written by Simon Anders.

Value

Vector with variances.

See Also

vignette("Intro2Matrix", package="Matrix") CsparseMatrix-class

Examples

library(Matrix)
 mat <- as(matrix(rpois(900,1), ncol=3), "dgCMatrix")
 colVars_spm(mat)

Feature Plot

Description

Highlight gene expression data in a 2D-embedding (UMAP, tSNE, etc.).

Usage

feat(embedding, expression, legend_name = "Expression")

Arguments

embedding

A matrix/data.frame/tibble/... with exactly two columns. If colnames are missing, the axis will be named "Dim1" and "Dim2". Other classes than matrix/data.frame/tibble are possible, as long as data.frame(embedding)) produces a numeric data.frame.
expression

Numeric vector with expression values of the gene of interest. Order has to correspond to the row order in embedding. Typically, expression is normalized gene expression and we recommend k/s/mean(1/s), where k are UMI counts for the gene of interest and s are totalUMI of the cell (aka library size).
legend_name

Text displayed above the legend. Most commonly the name of the displayed gene.

Details

This function discourages customization on purpose, because it bundles geoms, themes and settings that I found important for visualizing gene expression in scRNAseq data:

  • coord_fixed, to avoid distortion of embeddings

  • geom_point with size=.4, to ameliorate overplotting

  • No background grid, because distances and axis units in embeddings do not carry meaning for most dimensionality reduction techniques.

  • Intensity-coded color scales (viridis) displayed with log2-transformation. Makes visualization independent of colorblindness and appropriate for gene expression data (which is usually Log Normal distributed).

  • Color scale breaks are displayed as 'closed interval', i.e. max(expression) and min(expression) are the most extreme breaks. Rounding makes them human-readable. This functionality is provided by closed_breaks_log2 and closed_labels.

If you insist on customizing, think of this function as a great starting point, you can simply copy-paste the code after typing feat into your console.

Value

A ggplot2 object storing a colored scatter plot.

See Also

ggplot, closed_labels, closed_breaks_log2

Examples

# expression goes from 0 to 22:
 set.seed(100)
 feat(matrix(rnorm(2000, c(.1, 3)), ncol=2), rpois(1000, c(.1, 11)))
 # expression goes from 2 to 52:
 set.seed(100)
 feat(matrix(rnorm(2000, c(.1, 3)), ncol=2), rpois(1000, c(10, 31)))

Check if number(s) is/are integers. In contrast to is.integer, is_wholenumber does not check the class but accepts all numbers that are integers with reasonable precision.

Description

Check if number(s) is/are integers. In contrast to is.integer, is_wholenumber does not check the class but accepts all numbers that are integers with reasonable precision.

Usage

is_wholenumber(x, tol = .Machine$double.eps^0.5)

Arguments

x

Number to test
tol

tolerance for testing