scProkaR

scProkaR is a SingleCellExperiment-centred toolkit for microbial single-cell RNA sequencing. It standardises bacterial input objects, computes prokaryote-specific quality-control metrics such as ribosomal RNA fraction, integrates batches with several backends, and scores the corrected embeddings with scIB-inspired benchmarking metrics. It also provides pseudobulk differential expression built on edgeR, and Time Aware Trajectory Analysis (TATA), a time-informed trajectory abstraction for perturbation time courses.

Installation

scProkaR is under review for Bioconductor. Once it is accepted, install it with:

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
BiocManager::install("scProkaR")

To install the development version from GitHub:

BiocManager::install("Monash-Li-Lab/scProkaR")

Getting started

library(scProkaR)
data("sce1", package = "scProkaR")
sce1

The package ships sce1, a bacterial single-cell time course of 5000 cells across an untreated control and two polymyxin B concentrations, sampled at 0, 1, 4 and 7 hours.

A minimal workflow:

sce <- sce1
sce$batch <- factor(sce$treatments)
sce <- RunBacQC(sce)
sce <- FilterBacCells(sce, min_counts = 200, min_features = 50)

sce <- IntegrateBacData(sce, batch_col = "batch", method = "mnn", dims = 1:10)
sce <- RunIntegratedUMAP(sce, reduction = "integrated_mnn")
sce <- RunIntegratedClustering(sce, reduction = "integrated_mnn")

PlotReduction(sce, reduction = "umap_integrated_mnn", colour_by = "batch")

Time-aware trajectory analysis on the integrated embedding:

set.seed(101)
tata <- run_tata(
    sce,
    dimred = "integrated_mnn",
    time_col = "timepoints",
    cluster_col = "tata_cluster",
    k = 20
)
plot_tata_cluster_graph(tata)

Vignettes

Two vignettes document the package in full:

browseVignettes("scProkaR")

• scProkaR core workflow covers object creation, quality control, integration, benchmarking, clustering and pseudobulk differential expression.
• Time Aware Trajectory Analysis covers the TATA graph, pseudotime, branch probabilities and gene trends.

Main functions

Stage	Functions
Object creation	CreateBacObject(), MergeBacObjects()
Quality control	RunBacQC(), FilterBacCells()
Integration	IntegrateBacData(), RegisterIntegrationEmbedding(), RunIntegratedUMAP(), RunIntegratedClustering()
Benchmarking	BenchmarkIntegration()
Visualisation	PlotReduction(), PlotIntegrationOverview()
Pseudobulk DE	aggregate_pseudobulk(), normalize_pseudobulk(), run_edger_pairwise_de(), run_edger_spline_de(), run_scproka_de()
Trajectory	run_tata(), compute_tata_pseudotime(), compute_tata_branch_probabilities(), compute_tata_gene_trends()

Optional backends

Several backends are optional and are only needed for specific steps:

• batchelor for mutual nearest neighbour integration
• harmony for Harmony integration
• uwot for UMAP
• SeuratObject for reading Seurat objects into CreateBacObject()
• DropletUtils for reading 10x Genomics directories
• scrapper, irlba, BiocSingular, BiocNeighbors, RANN, FNN, cluster for individual preprocessing and neighbour-search steps

Getting help

Please open an issue at https://github.com/Monash-Li-Lab/scProkaR/issues, or ask on the Bioconductor support site using the scProkaR tag.

License

MIT. See LICENSE.