SCORPIUS

SCORPIUS an unsupervised approach for inferring linear developmental chronologies from single-cell RNA sequencing data. In comparison to similar approaches, it has three main advantages:

• It accurately reconstructs linear dynamic processes. The performance was evaluated using a quantitative evaluation pipeline and ten single-cell RNA sequencing datasets.

• It automatically identifies marker genes, speeding up knowledge discovery.

• It is fully unsupervised. Prior knowledge of the relevant marker genes or cellular states of individual cells is not required.

News:

• See news(package = "SCORPIUS") for a full list of changes to the package.

• Our preprint is on bioRxiv (Cannoodt et al. 2016).

• Check out our review on Trajectory Inference methods (Saelens et al. 2019).

Installing SCORPIUS

You can install:

• the latest released version from CRAN with

  install.packages("SCORPIUS")

• the latest development version from GitHub with

  devtools::install_github("rcannood/SCORPIUS", build_vignettes = TRUE)

If you encounter a bug, please file a minimal reproducible example on the issues page.

Learning SCORPIUS

To get started, read the introductory example below, or read one of the vignettes containing more elaborate examples:

• Running SCORPIUS on an AnnData object:
  vignette("anndata", package="SCORPIUS")
• Investigating dendritic cell maturation in dendritic cell progenitors:
  vignette("ginhoux", package="SCORPIUS")
• Running SCORPIUS on a Seurat object:
  vignette("seurat", package="SCORPIUS")
• Trajectory inference from simulated data:
  vignette("simulated-data", package="SCORPIUS")

Introductory example

This section describes the main workflow of SCORPIUS without going in depth in the R code. For a more detailed explanation, see the vignettes listed below.

To start using SCORPIUS, simply write:

library(SCORPIUS)

The ginhoux dataset (Schlitzer et al. 2015) contains 248 dendritic cell progenitors in one of three cellular cellular states: MDP, CDP or PreDC. Note that this is a reduced version of the dataset, for packaging reasons. See ?ginhoux for more info.

data(ginhoux)
expression <- ginhoux$expression
group_name <- ginhoux$sample_info$group_name

With the following code, SCORPIUS reduces the dimensionality of the dataset and provides a visual overview of the dataset. In this plot, cells that are similar in terms of expression values will be placed closer together than cells with dissimilar expression values.

space <- reduce_dimensionality(expression, "spearman")
draw_trajectory_plot(space, group_name, contour = TRUE)

To infer and visualise a trajectory through these cells, run:

traj <- infer_trajectory(space)
draw_trajectory_plot(space, group_name, traj$path, contour = TRUE)

To identify candidate marker genes, run:

# warning: setting num_permutations to 10 requires a long time (~30min) to run!
# set it to 0 and define a manual cutoff for the genes (e.g. top 200) for a much shorter execution time.
gimp <- gene_importances(
  expression, 
  traj$time, 
  num_permutations = 10, 
  num_threads = 8, 
  ntree = 10000,
  ntree_perm = 1000
) 

To select the most important genes and scale its expression, run:

gimp$qvalue <- p.adjust(gimp$pvalue, "BH", length(gimp$pvalue))
gene_sel <- gimp$gene[gimp$qvalue < .05]
expr_sel <- scale_quantile(expression[,gene_sel])

To visualise the expression of the selected genes, use the draw_trajectory_heatmap function.

draw_trajectory_heatmap(expr_sel, traj$time, group_name)

Finally, these genes can also be grouped into modules as follows:

modules <- extract_modules(scale_quantile(expr_sel), traj$time, verbose = F)
draw_trajectory_heatmap(expr_sel, traj$time, group_name, modules)

References