Identification of novel cell subtypes is critically crucial in revealing the pathogenesis and heterogeneity of disease, which provides unprecedented insights into the development of therapeutic strategies. Although numerous cell-type identification methods exist, these methods heavily rely on the reference with the fixed cell labels, which fail to uncover new cell subtypes marked with phenotypic molecules within a specific disease context. To fill this gap, we propose a pioneering reference-free annotation method, Subtypist, to identify disease-associated cell subtypes expressing phenotypic features using an ensemble-based strategy.
- install dependent packages
devtools
> install.packages(pkgs = 'devtools')- then install Subtypist
> devtools::install_github('ZJUFanLab/Subtypist')
# or download the repository as ZIP
> devtools::install_local("/path/to/Subtypist-main.zip")- Loading or subsetting specific cell type
# Load Seurat object from a full dataset and subset specific cell types
> FullObject <- Load("FullObject.RData")
> Seu <- subset(FullObject, idents = c("T cells")
# Alternatively, load a pre-processed Seurat object with annotated cell types, derived by subsetting the full dataset
> # Seu <- readRDS("Seu.rds")- Cell type identification using an ensemble strategy without reference
# Object: a Loading or subsetting specific cell type Seurat object
# min.resolution: the minimum value of resolution
# max.resolution: the maxium value of resolution
# use.assay: Name of assay to use
# cluster.assay: Name of the assay in the Seurat object to use for clustering
> result <- Subtypist_merge(object=Seu,min.resolution=0.3,max.resolution=1.5,by=0.1,use.assay="RNA",cluster_assay = "RNA")
# Show results
> print(result)
$Object
An object of class Seurat
1000 features across 1000 samples within 1 assay
Active assay: RNA (1000 features, 985 variable features)
3 dimensional reductions calculated: pca, umap, tsne
$result.table
resolution merge_cluster initial_cluster molecular_phenotype Score
1 0.4 0 0 Gene7, Gene679, Gene990 2.0322883
2 0.4 1 1 Gene570, Gene807, Gene470 0.6984284
3 0.4 2 2 Gene871, Gene559, Gene247 3.3095874
4 0.4 3 3 Gene474, Gene746, Gene790 3.5678922
5 0.4 4 4 Gene776, Gene507, Gene323 2.9614237
6 1.1 0 0 Gene7, Gene679, Gene990 2.0341814
7 1.1 1 1 Gene871, Gene559, Gene247 3.3095874
8 1.1 2 2,5 Gene470, Gene807, Gene243 0.0000000
9 1.1 3 3 Gene474, Gene746, Gene577 3.4186799
10 1.1 4 4,7 Gene807, Gene801, Gene566 0.0000000
11 1.1 5 6 Gene776, Gene507, Gene323 2.9614237
- Evaluating clustering resolutions and annotating subtypes with specific phenotypic markers
# To evaluate and rank clustering resolutions based on their corresponding subtype identification results.
> sortScore(result$result.table) ## resolution = 0.4: highest
# A tibble: 5 × 2
resolution value
<dbl> <dbl>
1 0.1 1.62
2 0.2 1.85
3 0.4 2.51
4 1.1 1.95
5 1.2 1.81
> # Add the result to the object
> Seu <- AddSubtypist(result$Object,result.table=result$result.table,prefix='Subtypist')
> # To assign more specific phenotypic molecules to each subtype,
> # the `select_index` parameter can be used to specify which gene to select
> Seu <- Subtypist::AddSubtypist(result$Object,resolution=c(0.4),result.table=result$result.table,prefix = 'Subtypist',meta.prefix = 'phenotypic melocules_',value.suffix='+ B',select_index=c('0'=1,'1'=1,'2'=1,'3'=2,'4'=3))
> print(unique(Seu@meta.data['phenotypic melocules_0.4'])
> [1] "Gene570+ B" "Gene7+ B" "Gene746+ B" "Gene871+ B" "Gene323+ B"
- Visualize subtype-level distributions across dimensionality reduction space (e.g., UMAP) using Subtypist_Dimplot(). This function overlays the specified phenotypic molecules annotations—derived at selected clustering resolutions—onto the Seurat object. For example:
> p <- Subtypist::Subtypist_Dimplot(Seu,result.table = result$result.table,resolution = c(0.4,1.1), show = "molecular_phenotype_",prefix = 'Subtypist')Subtypist was developed by Yue Yao. Should you have any questions, please contact Yue Yao at yuey@zju.edu.cn