PLEASE NOTE: this package is unmaintaned and will soon be deprecated.

compassR

This package provides a specialized pipeline for the analysis and interpretation of cell-to-cell metabolic heterogeneity based on the single-cell metabolic reaction consistency matrix produced by the COMPASS algorithm. It also includes a suite of expressive utility functions for conducting statistical analyses building thereupon.

The presentation of the algorithm assumes a single-cell data set. However, you may choose to group cells together (e.g. via metacell or micropooling) to reduce computational overhead. You may also apply COMPASS to bulk transcriptome data sets (e.g. bulk RNA-seq or microarray data sets) of ample size.

Installation

1. Make sure you have installed the devtools package from CRAN.
2. Run devtools::install_github("YosefLab/compassR").

You can accomplish both of these steps by running the following R code.

# Install devtools from CRAN.
install.packages("devtools")

# Install compassR from YosefLab.
devtools::install_github("YosefLab/compassR")

Usage

In the following tutorial, we'll explore the Th17 cell data set (Wagner et al.; Wang et al.) that ships with the package. It will help you get acquainted with the basics, while skipping over some of the finer details; if you're an advanced user looking for the full documentation, please refer to the the wiki instead.

Loading your data

Our first step is to specify a few settings via a CompassSettings object.

compass_settings <- CompassSettings$new(
    user_data_directory = system.file("extdata", "Th17", package = "compassR"),
    cell_id_col_name = "cell_id",
    gene_id_col_name = "HGNC.symbol"
)

There are 3 important parameters.

• user_data_directory is the path to the directory that contains the data you want to analyze. This directory should include files named "cell_metadata.csv", "reactions.tsv", and "linear_gene_expression_matrix.tsv".
• cell_id_col_name is the column in "cell_metadata.csv" that uniquely identifies the cells in your data set.
• And finally, gene_id_col_name is the column in "gene_metadata.csv" that uniquely identifies the genes you're interested in. Note that you do not have to provide this file unless you're an advanced user. By default, analyses will just use the one included in the version of RECON2 that comes with the package -- in which case you can use "HGNC.symbol" for human genes or "MGI.symbol" for mouse genes.

Now we can load our data by creating a CompassData object.

compass_data <- CompassData$new(compass_settings)

This line may take a minute to run. Under the hood, it's postprocessing the results of the COMPASS algorithm and populating a few tables that we'll find useful for our analyses later on:

Table	Type	Description
reaction_consistencies	Data frame	Each row is a reaction and each column is a cell. reaction_consistencies[i, j] is the consitency (or "compatibility") between reaction i and cell j.
metareaction_consistencies	Data frame	Each row is a metareaction and each column is a cell. metareaction_consistencies[i, j] is the consistency (or "compatibility") between metareaction i and cell j.
metabolic_genes	Tibble	Each row describes a gene in terms of its ID and whether it's a metabolic gene.
gene_expression_statistics	Tibble	Each row describes a cell in terms of its ID, total expression, metabolic expression, and metabolic activity.
cell_metadata	Tibble	The cell metadata from cell_metadata.csv. In this example it's the Th17 cell data from the papers linked above.
gene_metadata	Tibble	The gene metadata from the metabolic model (RECON2, by default).
metabolite_metadata	Tibble	The metabolite metadata from the metabolic model (RECON2, by default).
reaction_metadata	Tibble	The reaction metadata from the metabolic model (RECON2, by default).
reaction_partitions	Tibble	Each row describes a reaction in terms of its ID, undirected ID, direction, and which metareaction (i.e. reaction group) it belongs to.

Note that all the metadata tables' fields are read as characters, and must manually be coerced into other data types if desired.

Exploring the statistical analysis suite

Now we're ready to start our analysis! We begin by making a CompassAnalyzer object.

compass_analyzer <- CompassAnalyzer$new(compass_settings)

With the CompassAnalyzer, it's easy to conduct statistical analyses. Let's do a Wilcoxon rank-sum test for whether each reaction achieves a higher consistency among Th17p cells or Th17n cells.

group_A_cell_ids <-
    compass_data$cell_metadata %>%
    filter(cell_type == "Th17p") %>%
    pull(cell_id)
group_B_cell_ids <-
    compass_data$cell_metadata %>%
    filter(cell_type == "Th17n") %>%
    pull(cell_id)
wilcoxon_results <- compass_analyzer$conduct_wilcoxon_test(
    compass_data$reaction_consistencies,
    group_A_cell_ids,
    group_B_cell_ids,
    for_metareactions = FALSE
)

We can use functions from the tidyverse to combine the results of our Wilcoxon test with the data we loaded earlier. Then, with just a little ggplot2, we can even reproduce figures 2(c) and 2(e) from the papers linked above!