1 Introduction

A common application of single-cell RNA sequencing (RNA-seq) data is to identify discrete cell types. To take advantage of the large collection of well-annotated scRNA-seq datasets, scClassify package implements a set of methods to perform accurate cell type classification based on ensemble learning and sample size calculation.

This vignette will provide an example showing how users can use a pretrained model of scClassify to predict cell types. A pretrained model is a scClassifyTrainModel object returned by train_scClassify(). A list of pretrained model can be found in https://sydneybiox.github.io/scClassify/index.html.

First, install scClassify, install BiocManager and use BiocManager::install to install scClassify package.

# installation of scClassify
if (!requireNamespace("BiocManager", quietly = TRUE)) {
  install.packages("BiocManager")
}
BiocManager::install("scClassify")

2 Setting up the data

We assume that you have log-transformed (size-factor normalized) matrices as query datasets, where each row refers to a gene and each column a cell. For demonstration purposes, we will take a subset of single-cell pancreas datasets from one independent study (Wang et al.).

library(scClassify)
data("scClassify_example")
wang_cellTypes <- scClassify_example$wang_cellTypes
exprsMat_wang_subset <- scClassify_example$exprsMat_wang_subset
exprsMat_wang_subset <- as(exprsMat_wang_subset, "dgCMatrix")

Here, we load our pretrained model using a subset of the Xin et al. human pancreas dataset as our reference data.

First, let us check basic information relating to our pretrained model.

data("trainClassExample_xin")
trainClassExample_xin
#> Class: scClassifyTrainModel 
#> Model name: training 
#> Feature selection methods: limma 
#> Number of cells in the training data: 674 
#> Number of cell types in the training data: 4

In this pretrained model, we have selected the genes based on Differential Expression using limma. To check the genes that are available in the pretrained model:

features(trainClassExample_xin)
#> [1] "limma"

We can also visualise the cell type tree of the reference data.

plotCellTypeTree(cellTypeTree(trainClassExample_xin))

3 Running scClassify

Next, we perform predict_scClassify with our pretrained model trainRes = trainClassExample to predict the cell types of our query data matrix exprsMat_wang_subset_sparse. Here, we used pearson and spearman as similarity metrics.

pred_res <- predict_scClassify(exprsMat_test = exprsMat_wang_subset,
                               trainRes = trainClassExample_xin,
                               cellTypes_test = wang_cellTypes,
                               algorithm = "WKNN",
                               features = c("limma"),
                               similarity = c("pearson", "spearman"),
                               prob_threshold = 0.7,
                               verbose = TRUE)
#> Performing unweighted ensemble learning... 
#> Using parameters: 
#> similarity  algorithm   features 
#>  "pearson"     "WKNN"    "limma" 
#> [1] "Using dynamic correlation cutoff..."
#> [1] "Using dynamic correlation cutoff..."
#> classify_res
#>                correct   correctly unassigned           intermediate 
#>            0.704590818            0.239520958            0.000000000 
#> incorrectly unassigned         error assigned          misclassified 
#>            0.000000000            0.051896208            0.003992016 
#> Using parameters: 
#> similarity  algorithm   features 
#> "spearman"     "WKNN"    "limma" 
#> [1] "Using dynamic correlation cutoff..."
#> [1] "Using dynamic correlation cutoff..."
#> classify_res
#>                correct   correctly unassigned           intermediate 
#>            0.702594810            0.013972056            0.000000000 
#> incorrectly unassigned         error assigned          misclassified 
#>            0.001996008            0.277445110            0.003992016 
#> weights for each base method: 
#> numeric(0)

Noted that the cellType_test is not a required input. For datasets with unknown labels, users can simply leave it as cellType_test = NULL.

Prediction results for pearson as the similarity metric:

table(pred_res$pearson_WKNN_limma$predRes, wang_cellTypes)
#>                   wang_cellTypes
#>                    acinar alpha beta delta ductal gamma stellate
#>   alpha                 0   206    0     0      0     2        0
#>   beta                  0     0  118     0      1     0        0
#>   beta_delta_gamma      0     0    0     0     25     0        0
#>   delta                 0     0    0    10      0     0        0
#>   gamma                 0     0    0     0      0    19        0
#>   unassigned            5     0    0     0     70     0       45

Prediction results for spearman as the similarity metric:

table(pred_res$spearman_WKNN_limma$predRes, wang_cellTypes)
#>                   wang_cellTypes
#>                    acinar alpha beta delta ductal gamma stellate
#>   alpha                 0   206    0     0      0     2        2
#>   beta                  2     0  118     0     29     0        6
#>   beta_delta_gamma      1     0    0     0     66     0       31
#>   delta                 0     0    0    10      0     0        2
#>   gamma                 0     0    0     0      0    18        0
#>   unassigned            2     0    0     0      1     1        4

4 Session Info

sessionInfo()
#> R version 4.0.0 (2020-04-24)
#> Platform: x86_64-pc-linux-gnu (64-bit)
#> Running under: Ubuntu 18.04.4 LTS
#> 
#> Matrix products: default
#> BLAS:   /home/biocbuild/bbs-3.11-bioc/R/lib/libRblas.so
#> LAPACK: /home/biocbuild/bbs-3.11-bioc/R/lib/libRlapack.so
#> 
#> 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       
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#> [1] scClassify_1.0.0 BiocStyle_2.16.0
#> 
#> loaded via a namespace (and not attached):
#>  [1] Biobase_2.48.0      viridis_0.5.1       mixtools_1.2.0     
#>  [4] tidyr_1.0.2         tidygraph_1.1.2     viridisLite_0.3.0  
#>  [7] splines_4.0.0       ggraph_2.0.2        RcppParallel_5.0.0 
#> [10] assertthat_0.2.1    statmod_1.4.34      BiocManager_1.30.10
#> [13] stats4_4.0.0        yaml_2.2.1          ggrepel_0.8.2      
#> [16] pillar_1.4.3        lattice_0.20-41     glue_1.4.0         
#> [19] limma_3.44.0        digest_0.6.25       polyclip_1.10-0    
#> [22] colorspace_1.4-1    htmltools_0.4.0     Matrix_1.2-18      
#> [25] pkgconfig_2.0.3     magick_2.3          bookdown_0.18      
#> [28] purrr_0.3.4         scales_1.1.0        tweenr_1.0.1       
#> [31] hopach_2.48.0       BiocParallel_1.22.0 ggforce_0.3.1      
#> [34] tibble_3.0.1        proxy_0.4-24        mgcv_1.8-31        
#> [37] farver_2.0.3        ggplot2_3.3.0       ellipsis_0.3.0     
#> [40] BiocGenerics_0.34.0 survival_3.1-12     magrittr_1.5       
#> [43] crayon_1.3.4        evaluate_0.14       nlme_3.1-147       
#> [46] MASS_7.3-51.6       segmented_1.1-0     tools_4.0.0        
#> [49] minpack.lm_1.2-1    lifecycle_0.2.0     stringr_1.4.0      
#> [52] S4Vectors_0.26.0    kernlab_0.9-29      munsell_0.5.0      
#> [55] cluster_2.1.0       compiler_4.0.0      proxyC_0.1.5       
#> [58] rlang_0.4.5         grid_4.0.0          igraph_1.2.5       
#> [61] labeling_0.3        rmarkdown_2.1       gtable_0.3.0       
#> [64] graphlayouts_0.7.0  R6_2.4.1            gridExtra_2.3      
#> [67] knitr_1.28          dplyr_0.8.5         stringi_1.4.6      
#> [70] parallel_4.0.0      Rcpp_1.0.4.6        vctrs_0.2.4        
#> [73] diptest_0.75-7      tidyselect_1.0.0    xfun_0.13