Contents

1 scDblFinder

1.1 Introduction

scDblFinder identifies doublets in single-cell RNAseq directly by creating artificial doublets and looking at their prevalence in the neighborhood of each cell. The rough logic is very similar to DoubletFinder, but it much simpler and more efficient. In a nutshell:

  • scDblFinder works directly on a reduced count matrix (using rank correlation on top expressed genes), making the detection independent of downstream processing. This avoids the need for all the Seurat pre-processing, making the detection faster and compatible with other analytic choices.
  • instead of creating doublets from random pairs of cells, scDblFinder first overclusters the cells and create cross-cluster doublets. It also uses meta-cells from each cluster to create triplets. This strategy avoids creating homotypic doublets and enables the detection of most doublets with much fewer artificial doublets.
  • while we also rely on the expected proportion of doublets to threshold the scores, we include a variability in the estimate of the doublet proportion (dbr.sd), and use the error rate of the real/artificial predicition in conjunction with the deviation in global doublet rate to set the threshold.

1.2 Installation

scDblFinder was developed under R 3.6. Install with:

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
BiocManager::install("scDblFinder")

Or, until the new bioconductor release:

BiocManager::install("plger/scDblFinder")

1.3 Usage

Given an object sce of class SingleCellExperiment:

library(scDblFinder)
sce <- scDblFinder(sce)

This will add the following columns to the colData of sce:

  • sce$scDblFinder.neighbors : the number of neighbors considered
  • sce$scDblFinder.ratio : the proportion of artificial doublets among the neighborhood (the higher, the more chances that the cell is a doublet)
  • sce$scDblFinder.score : a doublet score integrating the ratio in a probability of the cell being a doublet
  • sce$scDblFinder.class : the classification (doublet or singlet)

1.3.1 Multiple samples

If you have multiple samples (understood as different cell captures), then it is preferable to look for doublets separately for each sample. You can do this by simply providing a vector of the sample ids to the samples parameter of scDblFinder or, if these are stored in a column of colData, the name of the column. In this case, you might also consider multithreading it using the BPPARAM parameter. For example:

library(BiocParallel)
sce <- scDblFinder(sce, samples="sample_id", BPPARAM=MulticoreParam(3))
table(sce$scDblFinder.class)

1.3.2 Parameters

The important sets of parameters in scDblFinder refer respectively to the expected proportion of doublets, to the clustering, and to the number of artificial doublets used.

1.3.2.1 Expected proportion of doublets

The expected proportion of doublets has no impact on the score (the ratio above), but a very strong impact on where the threshold will be placed. It is specified through the dbr parameter and the dbr.sd parameter (the latter specifies the standard deviation of dbr, i.e. the uncertainty in the expected doublet rate). For 10x data, the more cells you capture the higher the chance of creating a doublet, and Chromium documentation indicates a doublet rate of roughly 1% per 1000 cells captures (so with 5000 cells, (0.01*5)*5000 = 250 doublets), and the default expected doublet rate will be set to this value (with a default standard deviation of 0.015). Note however that different protocols may create considerably more doublets, and that this should be updated accordingly.

1.3.2.2 Clustering

Since doublets are created across clusters, it is important that subpopulations are not misrepresented as belonging to the same cluster. For this reason, we rely on an over-clustering approach which is similar to scran’s quickCluster, but splits clusters above a certain size. This is implemented by scDblFinder’s overcluster function. By default, the maximum cluster size will be 1/20 of the number of cells. While this is reasonable for most datasets (i.e. all those used for benchmark), so many clusters might be unnecessary when the cell population has a simple structure, and more clusters might be needed in a very complex population (e.g. whole brain).

1.3.2.3 Number of artificial doublets

scDblFinder itself determines a reasonable number of artificial doublets to create on the basis of the size of the population and the number of clusters, but increasing this number can only increase the accuracy. If you increase the number above default settings, you might also consider increasing parameter k - i.e. the number of neighbors considered.



2 Comparison with other tools

To benchmark scDblFinder against alternatives, we used datasets in which cells from multiple individuals were mixed and their identity deconvoluted using SNPs (via demuxlet), which also enables the identification of doublets from different individuals.

The method is compared to:

2.1 Mixology10x3cl

Accuracy of the doublet detection in the mixology10x3cl dataset (a mixture of 3 cancer cell lines). All methods perform very well.

Figure 1: Accuracy of the doublet detection in the mixology10x3cl dataset (a mixture of 3 cancer cell lines)
All methods perform very well.

2.2 Mixology10x5cl

Accuracy of the doublet detection in the mixology10x5cl dataset (a mixture of 5 cancer cell lines).

Figure 2: Accuracy of the doublet detection in the mixology10x5cl dataset (a mixture of 5 cancer cell lines)

2.3 Demuxlet controls

Accuracy of the doublet detection in the demuxlet control (Batch 2) dataset (GSM2560248).

Figure 3: Accuracy of the doublet detection in the demuxlet control (Batch 2) dataset (GSM2560248)

2.4 Running time

## Warning: Removed 1 rows containing missing values (position_stack).
Running time for each method/dataset

Figure 4: Running time for each method/dataset

1 DoubletFinder failed on the mixology10x3cl dataset

Note that by far most of the running time of scDblFinder is actually the clustering.

Session info

## R version 3.6.1 (2019-07-05)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 18.04.3 LTS
## 
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.10-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.10-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] scDblFinder_1.0.0 cowplot_1.0.0     ggplot2_3.2.1     BiocStyle_2.14.0 
## 
## loaded via a namespace (and not attached):
##  [1] Biobase_2.46.0              viridis_0.5.1              
##  [3] edgeR_3.28.0                BiocSingular_1.2.0         
##  [5] viridisLite_0.3.0           DelayedMatrixStats_1.8.0   
##  [7] assertthat_0.2.1            statmod_1.4.32             
##  [9] BiocManager_1.30.9          highr_0.8                  
## [11] stats4_3.6.1                dqrng_0.2.1                
## [13] GenomeInfoDbData_1.2.2      vipor_0.4.5                
## [15] yaml_2.2.0                  pillar_1.4.2               
## [17] lattice_0.20-38             glue_1.3.1                 
## [19] limma_3.42.0                digest_0.6.22              
## [21] GenomicRanges_1.38.0        XVector_0.26.0             
## [23] randomForest_4.6-14         colorspace_1.4-1           
## [25] htmltools_0.4.0             Matrix_1.2-17              
## [27] pkgconfig_2.0.3             bookdown_0.14              
## [29] zlibbioc_1.32.0             purrr_0.3.3                
## [31] scales_1.0.0                BiocParallel_1.20.0        
## [33] tibble_2.1.3                IRanges_2.20.0             
## [35] withr_2.1.2                 SummarizedExperiment_1.16.0
## [37] BiocGenerics_0.32.0         lazyeval_0.2.2             
## [39] magrittr_1.5                crayon_1.3.4               
## [41] evaluate_0.14               beeswarm_0.2.3             
## [43] tools_3.6.1                 scater_1.14.0              
## [45] data.table_1.12.6           matrixStats_0.55.0         
## [47] stringr_1.4.0               S4Vectors_0.24.0           
## [49] munsell_0.5.0               locfit_1.5-9.1             
## [51] DelayedArray_0.12.0         irlba_2.3.3                
## [53] compiler_3.6.1              GenomeInfoDb_1.22.0        
## [55] rsvd_1.0.2                  rlang_0.4.1                
## [57] grid_3.6.1                  RCurl_1.95-4.12            
## [59] BiocNeighbors_1.4.0         SingleCellExperiment_1.8.0 
## [61] igraph_1.2.4.1              bitops_1.0-6               
## [63] labeling_0.3                rmarkdown_1.16             
## [65] gtable_0.3.0                R6_2.4.0                   
## [67] gridExtra_2.3               knitr_1.25                 
## [69] dplyr_0.8.3                 stringi_1.4.3              
## [71] ggbeeswarm_0.6.0            parallel_3.6.1             
## [73] Rcpp_1.0.2                  scran_1.14.0               
## [75] tidyselect_0.2.5            xfun_0.10