About the tool

ScInfeR: a graph-based cell type annotation toolkit for single-cell RNA-seq, ATAC-seq, and spatial omics

Brief overview of the tool algorithm

1. ScInfeR can annotate cells using user-defined marker sets, scRNA-seq references, or both to annotate cells in scRNA-seq, scATAC-seq, and spatial omics datasets.
2. ScInfeR implements two rounds of annotation strategy for cell-type assignment.
3. First, the tool annotates the cell clusters by correlating the cluster-specific markers with the cell-type-specific markers in the cell-cell similarity graph. These cell-type-specific marker genes can be either user-defined, extracted by ScInfeR from scRNA-seq reference data, or a combination of both. For scRNA-seq as a reference, ScInfeR extracts cell-type markers by considering both the global and local specificity of markers.
4. In the second round, the tool annotates the subtypes and clusters containing multiple cell types in a hierarchical manner. In this step, the tool uses a framework adapted from the message-passing layer in the graph neural network to annotate each cell individually.

How to install

using devtools

devtools::install_github("swainasish/ScInfeR")

Or you can load the R script as source

source("https://raw.githubusercontent.com/swainasish/ScInfeR/master/R/base.R")

📚 Cite the Paper

Asish Kumar Swain, Rajveer Singh Shekhawat, Pankaj Yadav.
ScInfeR: an efficient method for annotating cell types and sub-types in single-cell RNA-seq, ATAC-seq, and spatial omics.
Briefings in Bioinformatics, Volume 26, Issue 3, May 2025, bbaf253.
https://doi.org/10.1093/bib/bbaf253

Tutorials

Annotate scRNA-seq datasets using marker-set

Load the libraries

library(Seurat)
library(ScInfeR)

Load the dataset and pre-process using Seurat

ts_pancreas <-  LoadSeuratRds("fig2_ts_pancreas.rds")
ts_pancreas  <- NormalizeData(ts_pancreas , normalization.method = "LogNormalize", scale.factor = 10000)
ts_pancreas  <- FindVariableFeatures(ts_pancreas , selection.method = "vst", nfeatures = 2000)
genes <- VariableFeatures(ts_pancreas )
ts_pancreas <- ScaleData(ts_pancreas , features = genes)
ts_pancreas  <- RunPCA(ts_pancreas , features = genes)
ts_pancreas  <- FindNeighbors(ts_pancreas , dims = 1:10)
ts_pancreas  <- FindClusters(ts_pancreas , resolution = 1)
ts_pancreas  <- RunUMAP(ts_pancreas , dims = 1:10)
ts_pancreas$"Ground truth" <- ts_pancreas$Celltype
DimPlot(ts_pancreas,reduction = "umap",group.by = "Ground truth")

Fetch the marker-set from ScInfeRDB
You can change the tissue type by change the argument

scinfer_marker_pancreas <- fetch_markerset(tissue_type = "Pancreas")

Celltype annotation by ScInfeR, using marker-set as guidance

scInfer_pred_pancreas_mar <- predict_celltype_scRNA_seurat(s_object = ts_pancreas ,
                                                       group_annt = ts_pancreas$seurat_clusters,
                                                       ct_marker_df = scinfer_marker_pancreas,
                                                       subtype_present = F,
                                                       subtype_info = F,
                                                       assay_name= "RNA",
                                                       slot_name = "counts",
                                                       own_weightage = 0.5,
                                                       n_neighbor=10)
ts_pancreas$ScInfeR <- scInfer_pred_pancreas_mar$celltype
scinfer_plot <- DimPlot(ts_pancreas,reduction = "umap",group.by = "ScInfeR")
gt+scinfer_plot

Annotate scRNA-seq datasets using scRNA-seq reference or hybrid mode

library(Seurat)
library(ScInfeR)
library(ggplot2)

Load the target dataset and pre-process

ts_pancreas <-  LoadSeuratRds("fig2_ts_pancreas.rds")
ts_pancreas  <- NormalizeData(ts_pancreas , normalization.method = "LogNormalize", scale.factor = 10000)
ts_pancreas  <- FindVariableFeatures(ts_pancreas , selection.method = "vst", nfeatures = 2000)
genes <- VariableFeatures(ts_pancreas )
ts_pancreas <- ScaleData(ts_pancreas , features = genes)
ts_pancreas  <- RunPCA(ts_pancreas , features = genes)
ts_pancreas  <- FindNeighbors(ts_pancreas , dims = 1:10)
ts_pancreas  <- FindClusters(ts_pancreas , resolution = 1)
ts_pancreas  <- RunUMAP(ts_pancreas , dims = 1:10)
ts_pancreas$"Ground truth" <- ts_pancreas$Celltype
gt <- DimPlot(ts_pancreas,reduction = "umap",group.by = "Ground truth")
gt

Load the reference dataset and fetch markers from the reference dataset
combined the markers fetched from reference dataset and markers from ScInfeRDB

scinfer_marker_pancreas <- fetch_markerset(tissue_type = "Pancreas")

scinferdb_pancreas <- fetch_scRNA_reference(tissue_type="Pancreas",path=F)
pancreas_marker <- get_marker_from_ref_matrix(exp_mat=scinferdb_pancreas@assays$RNA$counts,
                                           annotations=scinferdb_pancreas$Celltype,
                                           umap_cor=scinferdb_pancreas@[email protected],
                                           num_marker_per_ct=10)  
pancreas_marker_combined <- rbind(scinfer_marker_pancreas,pancreas_marker)

Predict the cell types

scInfer_pred_pancreas_ref <- predict_celltype_scRNA_seurat(s_object = ts_pancreas ,
                                                           group_annt = ts_pancreas$seurat_clusters,
                                                           ct_marker_df =pancreas_marker_combined  ,
                                                           subtype_present = F,
                                                           subtype_info = F,
                                                           assay_name= "RNA",
                                                           slot_name = "counts",
                                                           own_weightage = 0.5,
                                                           n_neighbor=10)
ts_pancreas$"scinfer" <- scInfer_pred_pancreas_ref$celltype
scinfer_plot <- DimPlot(ts_pancreas,reduction = "umap",group.by = "scinfer")
gt+scinfer_plot

Annotate scATAC-seq datsets

Load the libraries

library(Seurat)
library(ScInfeR)
library(ggplot2)

In the first part of this tutorial shows to annotate scATAC-seq dataset using scRNA-seq as reference
in next part we annotate the same dataset using scATAC-seq as reference.

Annotate scATAC-seq dataset using scRNA-seq as reference

Here, Arch object and gene activity score are needed for the analysis.

atac_GSE129785_arch <- readRDS("/scATAC/final_arch_objects/GSE129785.arch.rds")
atac_GSE129785_gene_activity <- LoadSeuratRds("/reproduce/fig3_gse129785_geneactivity.rds")

Next, the reference scRNA-seq dataset have to load

disco_ref_atac_pbmc <- LoadSeuratRds("/reproduce/fig3_disco_pbmc_atac_ref.rds")

annotate using ScInfeR
In first step, quality markers fetched from the reference scRNA-seq, next annotation can be performed.

pbmc_atac_marker <- get_marker_from_ref_matrix(exp_mat=disco_ref_atac_pbmc@assays$RNA$counts,
                                               annotations=disco_ref_atac_pbmc$Celltype, 
                                         umap_cor=disco_ref_atac_pbmc@[email protected],
                                               num_marker_per_ct=10)

scInfer_pred_atac_ref<- predict_celltype_scATAC(gene_act_mat=atac_GSE129785_gene_activity@assays$RNA$counts,
                                                  group_annt=atac_GSE129785_arch$Clusters,
                                                  ct_marker_df=pbmc_atac_marker,
                                                  umap_cord=atac_GSE129785_arch@embeddings$UMAP$df)

plot

atac_GSE129785_gene_activity@reductions$"umap" <- CreateDimReducObject(as.matrix(atac_GSE129785_arch@embeddings$UMAP$df),key = "umap",assay = "RNA")
atac_GSE129785_gene_activity$"Ground truth" <- atac_GSE129785_gene_activity$Celltype
atac_GSE129785_gene_activity$"ScInfeR" <- scInfer_pred_atac_ref$celltype
gt_plot <- DimPlot(atac_GSE129785_gene_activity,reduction = "umap",group.by = "Ground truth")
scinfer_plot <- DimPlot(atac_GSE129785_gene_activity,reduction = "umap",group.by = "ScInfeR")
gt_plot + scinfer_plot

Annotate scATAC-seq dataset using scATAC-seq as reference

Here quality markers are retrived from the scATAC-seq reference, as the tissue type is PBMC where cell types are highly similar to each other, n_local set at 2 and higher weight assigned to local_weightage.

atac_ref_arch <- readRDS("/scATAC/final_arch_objects/GSE123578.arch.rds")
atac_ref_gene_activity <- LoadSeuratRds("/reproduce/fig3_gse123578_geneacivity.rds")
atac_marker_123578 <- get_marker_from_ref_matrix(exp_mat=atac_ref_gene_activity@assays$RNA$counts,
                                              annotations=atac_ref_gene_activity$Celltype,
                                              umap_cor=atac_ref_arch@embeddings$UMAP$df,
                                              num_marker_per_ct=10,
                                              n_local = 2,
                                              auc_threshold=0.65,
                                              Local_weightage = 0.8) 

cell type annotaion

scInfer_pred_atac_ref2<- predict_celltype_scATAC(gene_act_mat=atac_GSE129785_gene_activity@assays$RNA$counts,
                                                  group_annt=atac_GSE129785_arch$Clusters,
                                                  ct_marker_df=atac_marker_123578,
                                                  umap_cord=atac_GSE129785_arch@embeddings$UMAP$df)

plot

atac_GSE129785_gene_activity$"Ground truth" <- atac_GSE129785_gene_activity$Celltype
atac_GSE129785_gene_activity$"ScInfeR" <- scInfer_pred_atac_ref2$celltype
gt_plot <- DimPlot(atac_GSE129785_gene_activity,reduction = "umap",group.by = "Ground truth")
scinfer_plot <- DimPlot(atac_GSE129785_gene_activity,reduction = "umap",group.by = "ScInfeR")
gt_plot + scinfer_plot

Annotate spatial omics datasets

load the libraries

library(Seurat)
library(ScInfeR)
library(ggplot2)

Load the datasets

starmap_scrna <- LoadSeuratRds("fig4a_starmap_scrna.rds" )
starmap_spatial <- LoadSeuratRds("fig4a_starmap_spatial.rds")

Here, instead of seurat cluster we will use spatial domains obtained from our recently published tool SpatialPrompt
In this case, user can user seurat clusters or the domains obtained from other tools also.
Article link for SpatialPrompt: https://doi.org/10.1038/s42003-024-06349-5.

spatial_prompt_domains <- read.csv("fig4a_starmap_sprompt_domains.csv",
                                   row.names = 1)

Fetch quality markers from scRNA-seq reference

common_genes = intersect(rownames(starmap_scrna),rownames(starmap_spatial))
starmap_scrna1 = starmap_scrna[common_genes,]
spatial_marker <- get_marker_from_ref_matrix(exp_mat=GetAssayData(starmap_scrna1),
                                             annotations=starmap_scrna1$celltype,
                                             umap_cor=starmap_scrna@[email protected],
                                             num_marker_per_ct=10)

Celltype prediction

scInfer_pred_starmap <- predict_celltype_spatial(expression_matrix =GetAssayData(starmap_spatial),
                                        group_annt=spatial_prompt_domains$X0,
                                        ct_marker_df=spatial_marker,
                                        spatial_cord =starmap_spatial@[email protected],
                                        subtype_present = F,
                                        subtype_info = F,
                                        own_weightage = 0.5,
                                        n_neighbor=10)

Plots

starmap_spatial$"Ground truth" <- starmap_spatial$celltype
starmap_spatial$"scinfer" <- scInfer_pred_starmap$celltype
plot <- DimPlot(starmap_spatial,reduction = "spatial",group.by = c("Ground truth","scinfer"))
plot