Wrappers for Scater v1.8.4

This code wraps a number of scater functions, ultimately for their use in Galaxy wrappers. Each script can be used in isolation. Briefly, the scater-create-qcmetric-ready-sce.R script takes a sample x gene expression matrix (usually read-counts) and cell annotation file, creates a SingleCellExperiment object and runs scater's calculateQCMetrics function (using other supplied files such as ERCC's and mitochondrial gene features). Various filter scripts are provided, along with some plotting functions for QC.

Install

The scripts can be used direclty from the command line, although the following R packages are required.

>source("https://bioconductor.org/biocLite.R")
Bioconductor version 3.7 (BiocInstaller 1.30.0), ?biocLite for help
>biocLite("scater")           #Should install >= scater-1.8.4
>devtools::install_github("ebi-gene-expression-group/workflow-scripts-common")
>install.packages(c("ggpubr", "optparse", "mvoutlier"))

R-3.5.1 was used to develop the code. Earlier versions have not been tested.

Commands

For help with any of the following scripts, run: <script-name> --help

---

scater-create-qcmetric-ready-sce.R Takes an expression matrix (usually read-counts) of samples (columns) and gene/transcript features (rows), along with other annotation information, such as cell metadata, control genes (mitochondrail genes, ERCC's), creates a SingleCellExperiment object and runs scater's calculateQCMetrics. Returns a SingleCellExperiment RDS object.

./scater-create-qcmetric-ready-sce.R -a test-data/counts.txt -c test-data/annotation.txt -f test-data/mt_controls.txt  -o test-data/scater_qcready_sce.rds

---

scater-plot-dist-scatter.R Takes SingleCellExperiment object (from RDS file) and lots a panel of read and feature graphs, including the distribution of library sizes, distribution of feature counts, a scatterplot of reads vs features, and % of mitochondrial genes in library.

./scater-plot-dist-scatter.R -i test-data/scater_qcready_sce.rds -o test-data/scater_reads_genes_dist.pdf

scater-plot-exprs-freq.R Plots mean expression vs % of expressing cells and provides information as to the number of genes expressed in 50% and 25% of cells.

scater-pca-filter.R Takes SingleCellExperiment object (from RDS file) and automatically removes outliers from data using PCA. Returns a filtered SingleCellExperiment RDS object.

./scater-pca-filter.R -i test-data/scater_qcready_sce.rds -o test-data/scater_pca_filtered.rds

---

scater-manual-filter.R Takes SingleCellExperiment object (from RDS file) and filters data using user-provided parameters. Returns a filtered SingleCellExperiment RDS object.

./scater-manual-filter.R -i test-data/scater_qcready_sce.rds -l 10000 -d 4 -m 33 -o test-data/scater_manual_filtered.rds

---

scater-normalize.R Compute log-normalized expression values from count data in a SingleCellExperiment object, using the size factors stored in the object. Returns a normalised SingleCellExperiment RDS object.

./scater-normalize.R -i test-data/scater_manual_filtered.rds -o test-data/scater_man_filtered_normalised.rds

---

scater-plot-pca.R PCA plot of a normalised SingleCellExperiment object (use scater-normalize.R before use). The options -c, -p, and -s all refer to cell annotation features. These are the column headers of the -c option used in scater-create-qcmetric-ready-sce.R.

./scater-plot-pca.R -i test-data/scater_man_filtered_normalised.rds -c Treatment -p Mutation_Status -o test-data/scater_pca_plot.pdf

Typical workflow

1. Read in data

./scater-create-qcmetric-ready-sce.R -a test-data/counts.txt -c test-data/annotation.txt -f test-data/mt_controls.txt  -o test-data/scater_qcready_sce.rds

2. Visualise it

Take a look at the distribution of library sizes, expressed features and mitochondrial genes.

./scater-plot-dist-scatter.R -i test-data/scater_qcready_sce.rds -o test-data/scater_reads_genes_dist.pdf

Then look at the distibution of genes across cells

./scater-plot-exprs-freq.R -i test-data/scater_qcready_sce.rds -o test-data/scater_exprs_freq.pdf

3. Use the plots to decide on filtering parameters

Option 1. Automatically filter on outliers

./scater-pca-filter.R -i test-data/scater_qcready_sce.rds -o test-data/scater_filtered.rds

Option 2. Manually filter

./scater-manual-filter.R -i test-data/scater_qcready_sce.rds -l 10000 -d 4 -m 33 -o test-data/scater_filtered.rds

4. Visualise data again to see how the filtering performed

./scater-plot-dist-scatter.R -i test-data/scater_filtered.rds -o test-data/scater_filtered_genes_dist.pdf

Decide if you're happy with the data. If not, try increasing or decreasing the filtering parameters.

5. Investigate other confounding factors

Plot the data (using PCA) and display various annotated properties of the cells

./scater-plot-pca.R -i test-data/scater_filtered_normalised.rds -c Treatment -p Mutation_Status -o test-data/scater_pca_plot.pdf