STACAS is a method for scRNA-seq integration, specifically designed to align scRNA-seq datasets that are composed of only partially overlapping cell populations or sub-types. It is based on the Seurat integration framework, and adds the following innovations:
- Anchors are down-weighted based on their distance in reciprocal PCA space
- Guide trees for pairwise integration are constructed based on the 'centrality' of datasets, as measured by the sum of their re-weighted anchor scores
- Prior knowledge, given as cell labels, can be used by the algorithm to remove inconsistent anchors, and perform semi-supervised integration
To install STACAS directly from the Git repository, run the following code from within RStudio:
if (!requireNamespace("remotes")) install.packages("remotes")
library(remotes)
remotes::install_github("carmonalab/STACAS")Standard integration (more here)
library(STACAS)
# get the test dataset "pbmcsca" from SeuratData package
if (!requireNamespace("remotes")) install.packages("remotes")
if (!requireNamespace("SeuratData")) install_github('satijalab/seurat-data')
library(SeuratData)
library(Seurat)
InstallData("pbmcsca")
data("pbmcsca")
# Integrate scRNA-seq datasets generated with different methods/technologies
pbmcsca.integrated <- NormalizeData(pbmcsca) |>
SplitObject(split.by = "Method")|>
Run.STACAS()
pbmcsca.integrated <- RunUMAP(pbmcsca.integrated, dims = 1:30)
# Visualize
DimPlot(pbmcsca.integrated, group.by = c("Method","CellType")) Semi-supervised integration (more here)
pbmcsca.semisup <- NormalizeData(pbmcsca) |>
SplitObject(split.by = "Method")|>
Run.STACAS(cell.labels = "CellType")
pbmcsca.semisup <- RunUMAP(pbmcsca.semisup, dims = 1:30) Find a tutorial for STACAS in a complete Seurat integration pipeline at: STACAS demo (code and instructions here)
See also how STACAS compares to Seurat for the integration of heterogeneos data sets: STACAS vs Seurat
Massimo Andreatta, Santiago J Carmona "STACAS: Sub-Type Anchor Correction for Alignment in Seurat to integrate single-cell RNA-seq data", Bioinformatics (2020) - https://doi.org/10.1093/bioinformatics/btaa755
