In samuel-marsh/scCustomize: Custom Visualizations & Functions for Streamlined Analyses of Single Cell Sequencing

---

all_times <- list()  # store the time for each chunk
knitr::knit_hooks$set(time_it = local({
  now <- NULL
  function(before, options) {
    if (before) {
      now <<- Sys.time()
    } else {
      res <- difftime(Sys.time(), now, units = "secs")
      all_times[[options$label]] <<- res
    }
  }
}))
knitr::opts_chunk$set(
  tidy = TRUE,
  tidy.opts = list(width.cutoff = 95),
  message = FALSE,
  warning = FALSE,
  time_it = TRUE
)

Interactivity with LIGER Objects

THIS VIGNETTE IS UNDER ACTIVE REVISION AND IS BEING UPDAATED WITH scCustomize V3.0.0 FUNCTIONS, BUT IS NOT YET COMPLETE.

While the majority of scCustomize's current functionality is with Seurat Objects there are a number of functions for working with LIGER objects too.

For this tutorial, I will be utilizing microglia data from Marsh et al., 2022 (Nature Neuroscience) which I have converted to LIGER object.

# Load Packages
library(ggplot2)
library(dplyr)
library(magrittr)
library(patchwork)
library(Seurat)
library(rliger)
library(scCustomize)
library(qs)

# read object
marsh_micro <- qread("assets/marsh_2020_micro.qs")

# Convert to liger
marsh_liger <- as.liger(object = marsh_micro, datasetVar = "orig.ident")

marsh_liger <- qread("assets/marsh_liger.qs")

LIGER object version

The Welch lab has recently released a major update to rliger package (v2.0.0+) which has substantial changes to object structure and major performance improvements over prior versions. The majority of liger functions in scCustomize v2.1.0 have been updated to work with old and new liger objects. However, a number of newer functions will be exclusive to rliger 2.0.0+.

This vignette will be using rliger 2.0.0.

Dimensionality Reduction Plotting

scCustomize contains new LIGER plotting function DimPlot_LIGER() that both adds functionality not present in LIGER package and provides function for plotting that is easier to use for those more familiar with Seurat's DimPlot()

DimPlot_LIGER(liger_object = marsh_liger)

As you can see we get plot that looks very similar to those generated by DimPlot_scCustom() or Seurat::DimPlot().

Optional functionality/Parameters

DimPlot_LIGER adds a number of options not present in liger plotting and modifies naming/functionality of others to be more in-line with Seurat for ease of use.

Plot combination

DimPlot_LIGER(liger_object = marsh_liger, combination = T)

User can specify different color palettes for each plot

clu_color <- ColorBlind_Pal()
sample_color <- DiscretePalette_scCustomize(num_colors = 12, palette = "alphabet")

DimPlot_LIGER(liger_object = marsh_liger, combination = T, colors_use_cluster = clu_color, colors_use_meta = sample_color)

Grouping and Splitting Plots

DimPlot_LIGER supports grouping and splitting plots by meta data variables or cluster and will return patchwork layout

DimPlot_LIGER(liger_object = marsh_liger, group_by = "Transcription_Method", colors_use_meta = JCO_Four())

DimPlot_LIGER(liger_object = marsh_liger, group_by = "Cluster", split_by = "Transcription_Method", label = FALSE)

Cluster Label Modification

DimPlot_LIGER supports many different modifications of cluster labels. NOTE: labeling is only supported when plotting by cluster.

• label logical. Whether to label clusters on plot.
• label_size Change label text size.
• label_repel logical. Whether to repel labels.
• label_box logical. Whether to place box around label.
• label_color change label text color.

DimPlot_LIGER(liger_object = marsh_liger)
DimPlot_LIGER(liger_object = marsh_liger, label = FALSE)
DimPlot_LIGER(liger_object = marsh_liger, label_repel = TRUE)
DimPlot_LIGER(liger_object = marsh_liger, label_repel = TRUE, label_box = TRUE, label_size = 3)

p1 <- DimPlot_LIGER(liger_object = marsh_liger)
p2 <- DimPlot_LIGER(liger_object = marsh_liger, label = FALSE)
p3 <- DimPlot_LIGER(liger_object = marsh_liger, label_repel = TRUE)
p4 <- DimPlot_LIGER(liger_object = marsh_liger, label_repel = TRUE, label_box = TRUE, label_size = 4)

wrap_plots(p1, p2, p3, p4, ncol = 2)

Raster Points

DimPlot_LIGER() also supports point rasterization utilizing scattermore package. Axes, labels, and legends will remain in vector form.

DimPlot_LIGER(liger_object = marsh_liger, raster = FALSE, label = FALSE)
DimPlot_LIGER(liger_object = marsh_liger, raster = TRUE, label = FALSE)

p6 <- DimPlot_LIGER(liger_object = marsh_liger, raster = FALSE, label = FALSE)
p7 <- DimPlot_LIGER(liger_object = marsh_liger, raster = TRUE, label = FALSE)

wrap_plots(p6, p7, ncol = 2)

Restore ggplot2 default color scheme

If you would like to restore ggplot2 default color scheme but maintain the rest of the aesthetics of DimPlot_LIGER simply set parameter ggplot_default_colors = TRUE.

DimPlot_LIGER(liger_object = marsh_liger, ggplot_default_colors = TRUE)

Customized plotFactors Functionality

NOTE: For this example as "marsh_liger" was not originally a LIGER analyzed object it has no factors. So I ran a basic LIGER pipeline (with no specific parameters to demonstrate functionality of Top_Genes_Factor)

marsh_liger <- qread("assets/marsh_liger_factor.qs")

scCustomize contains function plotFactors_scCustom() which expands the functionality and customizes the visualizes of LIGER's plotFactors() function.

Plots Created

plotFactors_scCustom returns two different plots for each iNMF factor. First, is plot of the factor loadings for each cell, grouped by dataset. The second is visualization of the factor loading on the dimensionality reduction coordinates.

returned_plots <- plotFactors_scCustom(liger_object = marsh_liger, return_plots = TRUE, save_plots = FALSE, raster = FALSE, order = TRUE)

p1 <- returned_plots[[1]][[6]]
p2 <- returned_plots[[2]][[6]]

wrap_plots(p1, p2, ncol = 2)

Additional Plot Save/Return Functionality

plotFactors_scCustom has a number of additional parameters that expand the functionality of plotFactors.

• save_plots and/or return_plots logical parameters to either return plots to assigned variable in environment or save the plots to PDF (or both).
• file_path and file_name optional parameters required if saving plots to PDF.
• raster whether or not to raster the points in both plots. Significantly reduced PDF file size and speed of plotting.

Additional Factor Plot Functionality

plotFactors_scCustom has a number of additional parameters that significantly expand the visualization options of plotFactors.

• colors_use_factors can be used to color the datasets in the factor loading plot. Must be vector equal to or greater than number of datasets in object. By default uses the "varibow" palette from DiscretePalette_scCustom instead of default ggplot2 hue palette.
• ggplot_default_colors parameter can be specified if you would like to return this plot to using the ggplot2/hue palette.

returned_plots2 <- plotFactors_scCustom(liger_object = marsh_liger, return_plots = TRUE, save_plots = FALSE, raster = FALSE, order = TRUE, ggplot_default_colors = TRUE)

p1 <- returned_plots[[1]][[6]]
p3 <- returned_plots2[[1]][[6]]

wrap_plots(p3, p1, ncol = 2)

Added Dimensionality Reduction Functionality

• colors_use_dimreduc can be used to define colors for dimensionality reduction plot. Can either be vector or two colors to be used to create gradient or a define gradient.
• order whether or not to reorder the points so that higher loading points are plotted on top of lower loading cells (avoids cell hiding). Default is FALSE.

/ IMPORTANT: AXES LABELS /

Unfortunately, the liger/rliger package does not properly store the name of the dimensionality reduction technique used and labels all plots with tSNE1/tSNE2 or dr1/dr2.

plotFactors_scCustom solves this by providing extra parameter: reduction_label in the function that can be used to set labels. The first time you use plotFactors_scCustom in R session it will print a one-time message alerting you to the fact that this parameter exists and is set to "UMAP" by default (as it seems to be most commonly used method currently).

Extract Top Loading Genes per Factor

Following plotting with plotFactors_scCustom it can sometimes be helpful to pull list of top loading genes for particular factor. Top_Genes_Factor provides quick function to extract the list of genes that loads highly on particular iNMF factor.
NOTE: This function simply extracts top loading genes in order from given factor. It does not assess whether that number of genes is significant, etc.

top20_factor6 <- Top_Genes_Factor(liger_object = marsh_liger, liger_factor = 6, num_genes = 20)

head(top20_factor6, 5)

Add Cell QC Metrics

For more information on adding cell QC metricss to LIGER objects see Object QC Vignette for information.

Variable Gene Selection

By default when selecting highly variable genes in LIGER using rliger::selectGenes() LIGER performs gene selection across all datasets in the object and then can either take the union or intersection of those lists (default is union). However, sometimes for analysis you may want to select highly variable genes from all data combined.

scCustomize provides the function Variable_Features_ALL_LIGER() that will do just that and returns the variable genes list to the @var.genes slot of original LIGER object.
NOTE: This function creates a temporary LIGER object by merging all of the matrices present in original object and then running variable gene selection. Therefore, while running the function it will use significant memory resources (equivalent to size of original object).

# Select highly variable genes based on threshold
marsh_liger <- Variable_Features_ALL_LIGER(liger_object = marsh_liger, var.thresh = 0.3)

# Select based on specific number of variable genes
marsh_liger <- Variable_Features_ALL_LIGER(liger_object = marsh_liger, num_genes = 2000)

Gene Utilities

As discussed in helpers vignette, the Gene_Present() function can be used to check if list of genes are present in a LIGER object.

By default this will query all datasets in the object to create unified gene list and check input gene list against that list.

gene_input_list <- c("P2ry12", "Fcrls", "Aif1")

genes_present <- Feature_Present(data = marsh_liger, features = gene_input_list)

However you can check individual datasets within object by supplying those matrices to the data parameter.

gene_input_list <- c("P2ry12", "Fcrls", "Aif1")

genes_present <- Feature_Present(data = marsh_liger@datasets[[1]], features = gene_input_list)

samuel-marsh/scCustomize documentation built on Dec. 20, 2024, 7:41 a.m.