Rank Constrained Similarity Learning for Single Cell RNA Sequencing Data

A novel clustering algorithm and toolkit to accurately identify various cell types using single cell RNA sequencing data from a complex tissue. This algorithm considers both local similarity and global similarity among the cells to discern the subtle differences among cells of the same type as well as larger differences among cells of different types. This algorithm uses Spearman’s rank correlations of a cell’s expression vector with those of other cells to measure its global similarity, and learns neighbour representation of a cell as its local similarity. The overall similarity of a cell to other cells is a linear combination of its global similarity and local similarity. See Mei et. al. (2021) for more details.

Rank Constrained Similarity Learning (RCSL)

RCSL is an R toolkit for single-cell clustering and trajectory analysis using single-cell RNA-seq data.

Installation

This package can be installed through devtools in R:

$ R
> library("devtools")
> devtools::install_github("QinglinMei/RCSL")

Now RCSL can be loaded in R:

> library(RCSL)

Input

The input of RCSL is a normalized data matrix with columns being cells and rows being genes in log(CPM+1), log(RPKM+1), log(TPM+1) or log(FPKM+1) format; or a data file in RDS format.

Usage

We provide an example script to run RCSL in demo_RCSL.R.

The nine functions of RCSL can also be run independently.

Example:

Load packages:

> library(RCSL)
> library(SingleCellExperiment)
> library(ggplot2)
> library(igraph)

Load Yan dataset:

> origData <- yan
> data <- logcounts(origData+1)
> label <- origData$cell_type1
> DataName <- "Yan"

Generating clustering result:

> res_RCSL <- RCSL(data)

Calculating Adjusted Rand Index:

> ARI_RCSL <- igraph::compare(res_RCSL$y, label, method = "adjusted.rand")

Trajectory analysis:

> label <- origData$cell_type1
> res_TrajecAnalysis <- TrajectoryAnalysis(res_RCSL$gfData, res_RCSL$drData, res_RCSL$S,
                                         clustRes = res_RCSL$y, TrueLabel = label, startPoint = 1,
                                         dataName = DataName)

Display the plot of constructed MST:

> res_TrajecAnalysis$MSTPlot

Display the plot of the pseudo-temporal ordering

> res_TrajecAnalysis$PseudoTimePlot

Display the plot of the inferred developmental trajectory

> res_TrajecAnalysis$TrajectoryPlot

A vignette in R Notebook format is available here

Required annotations for RCSL

1. The RCSL package requires three extra packages: namely the SingleCellExperiment package (see https://bioconductor.org/packages/release/bioc/html/SingleCellExperiment.html) to read the SingleCellExperiment object, the igraph package (see https://igraph.org/) to find the strongest connected components and the ggplot2 package (see https://cran.r-project.org/web/packages/ggplot2/index.html) to plot the developmental trajectory and MST.
2. The data for the demonstration purpose in the directory Data was from https://hemberg-lab.github.io/scRNA.seq.datasets/. This data is stored in both RDS and text formats.

DEBUG

Please feel free to contact us if you have problems running our tool at [email protected].