Cell type classification with SignacX: 1k PBMCs from 10X Genomics

This vignette shows how to use Signac with Seurat. There are three parts: Seurat, Signac and then visualization. We use an example PBMCs scRNA-seq data set from 10X Genomics.

Seurat

Start with the standard pre-processing steps for a Seurat object.

library(Seurat)

Download data from 10X Genomics.

dir.create("fls")
download.file("https://cf.10xgenomics.com/samples/cell-exp/3.0.0/pbmc_1k_v3/pbmc_1k_v3_filtered_feature_bc_matrix.h5",
    destfile = "fls/pbmc_1k_v3_filtered_feature_bc_matrix.h5")

Create a Seurat object, and then perform SCTransform normalization. Note:

You can use the legacy functions here (i.e., NormalizeData, ScaleData, etc.), use SCTransform or any other normalization method (including no normalization). We did not notice a significant difference in cell type annotations with different normalization methods.
We think that it is best practice to use SCTransform, but it is not a necessary step. Signac will work fine without it.

# load data
E = Read10X_h5(filename = "fls/pbmc_1k_v3_filtered_feature_bc_matrix.h5")
pbmc <- CreateSeuratObject(counts = E, project = "pbmc")

# run sctransform
pbmc <- SCTransform(pbmc, verbose = FALSE)

Perform dimensionality reduction by PCA and UMAP embedding. Note:

Signac actually needs these functions since it uses the nearest neighbor graph generated by Seurat.

# These are now standard steps in the Seurat workflow for visualization and clustering
pbmc <- RunPCA(pbmc, verbose = FALSE)
pbmc <- RunUMAP(pbmc, dims = 1:30, verbose = FALSE)
pbmc <- FindNeighbors(pbmc, dims = 1:30, verbose = FALSE)

SignacX

First, make sure you have the Signac package installed.

install.packages("SignacX")

Load the library

# load library
library(SignacX)

Generate SignacX labels for the Seurat object. Note:

Optionally, you can do parallel computing by setting num.cores > 1 in the Signac function.
Run time is ~10 minutes for ~10,000 cells.

# Run Signac
labels <- Signac(pbmc, num.cores = 4)
celltypes = GenerateLabels(labels, E = pbmc)

Sometimes, training the neural networks takes a lot of time. To make Signac faster, we implemented SignacFast which uses an ensemble of pre-trained neural network models. Note:

SignacFast uses an ensemble of 1,800 pre-calculated neural networks rather than performing bespoke model training.

# Run Signac
labels_fast <- SignacFast(pbmc)
celltypes_fast = GenerateLabels(labels_fast, E = pbmc)

SignacFast took only ~30 seconds. Relative to Signac, the main difference is that SignacFast tends to leave a few more cells “Unclassified.”

How does SignacFast compare to Signac?

	B	MPh	TNK	Unclassified
B	186	0	0	0
MPh	0	362	0	54
TNK	0	0	573	3
Unclassified	0	0	0	44

Visualizations

Now we can visualize the cell type classifications at many different levels: Immune and nonimmune

pbmc <- AddMetaData(pbmc, metadata = celltypes$Immune, col.name = "immmune")
pbmc <- SetIdent(pbmc, value = "immmune")
png(filename = "fls/plot1.png")
DimPlot(pbmc)
dev.off()

Immune, Nonimmune (if any) and unclassified cells

pbmc <- AddMetaData(pbmc, metadata = celltypes$L2, col.name = "L2")
pbmc <- SetIdent(pbmc, value = "L2")
png(filename = "fls/plot2.png")
DimPlot(pbmc)
dev.off()

Myeloid and lymphocytes

lbls = factor(celltypes$CellTypes)
levels(lbls) <- sort(unique(lbls))
pbmc <- AddMetaData(pbmc, metadata = lbls, col.name = "celltypes")
pbmc <- SetIdent(pbmc, value = "celltypes")
png(filename = "./fls/plot3.png")
DimPlot(pbmc)
dev.off()

Cell types

pbmc <- AddMetaData(pbmc, metadata = celltypes$CellTypes_novel, col.name = "celltypes_novel")
pbmc <- SetIdent(pbmc, value = "celltypes_novel")
png(filename = "./fls/plot4.png")
DimPlot(pbmc)
dev.off()

Cell types with novel populations

pbmc <- AddMetaData(pbmc, metadata = celltypes$CellStates, col.name = "cellstates")
pbmc <- SetIdent(pbmc, value = "cellstates")
png(filename = "./fls/plot5.png")
DimPlot(pbmc)
dev.off()

Cell states

Identify differentially expressed genes between cell types. Here, we see that Signac identified two novel cell populations that are positive for platelet and plasma cell markers, respectively.

pbmc <- SetIdent(pbmc, value = "celltypes_novel")

# Find protein markers for all clusters, and draw a heatmap
markers <- FindAllMarkers(pbmc, only.pos = TRUE, verbose = F, logfc.threshold = 1)
library(dplyr)
top5 <- markers %>%
    group_by(cluster) %>%
    top_n(n = 5, wt = avg_logFC)
png(filename = "./fls/plot6.png")
DoHeatmap(pbmc, features = unique(top5$gene), angle = 90)
dev.off()

Immune marker genes

Save results

saveRDS(pbmc, file = "fls/pbmcs_signac.rds")
saveRDS(celltypes, file = "fls/celltypes.rds")
saveRDS(celltypes_fast, file = "fls/celltypes_fast.rds")

Session Info

## R version 4.4.3 (2025-02-28)
## Platform: x86_64-pc-linux-gnu
## Running under: Ubuntu 24.04.2 LTS
## 
## Matrix products: default
## BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
## LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so;  LAPACK version 3.12.0
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## time zone: Etc/UTC
## tzcode source: system (glibc)
## 
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base     
## 
## loaded via a namespace (and not attached):
##  [1] digest_0.6.37     R6_2.6.1          fastmap_1.2.0     xfun_0.51        
##  [5] maketools_1.3.2   cachem_1.1.0      knitr_1.50        htmltools_0.5.8.1
##  [9] rmarkdown_2.29    buildtools_1.0.0  lifecycle_1.0.4   cli_3.6.4        
## [13] sass_0.4.9        jquerylib_0.1.4   compiler_4.4.3    sys_3.4.3        
## [17] tools_4.4.3       evaluate_1.0.3    bslib_0.9.0       yaml_2.3.10      
## [21] formatR_1.14      jsonlite_2.0.0    rlang_1.1.5