The cellustering is a toolkit designed for the preprocessing and clustering analysis of single-cell RNA sequencing (scRNA-seq) data.
You can install the development version of cellustering from GitHub with:
# install.packages("devtools")
devtools::install_github("langshangyuan/cellustering")This is a basic example which shows you how to create plots to facilitate quality control:
library(cellustering)
#>
#> Attaching package: 'cellustering'
#> The following object is masked from 'package:stats':
#>
#> kmeans
## basic example code
pbmc <- read_10x(cellustering_example("hg19"))
pbmc <- qc_plot(pbmc)
head(pbmc@quality$cell)
#> count_depth feature_count mito_percent
#> AAACATACAACCAC.1 2421 781 3.097893
#> AAACATTGAGCTAC.1 4903 1352 3.875178
#> AAACATTGATCAGC.1 3149 1131 1.079708
#> AAACCGTGCTTCCG.1 2639 960 1.970443
#> AAACCGTGTATGCG.1 981 522 1.427115
#> AAACGCACTGGTAC.1 2164 782 1.940850
head(pbmc@quality$feature)
#> total_expression cell_count
#> MIR1302.10 0 0
#> FAM138A 0 0
#> OR4F5 0 0
#> RP11.34P13.7 0 0
#> RP11.34P13.8 0 0
#> AL627309.1 9 9
pbmc@quality$cell_plot
pbmc@quality$violin_plot
pbmc@quality$feature_plot
After we perform the preprocessing steps like filter, normalize, we can
call the Cellustering package to find the High Variation Genes,
perform principal component analysis, and call a clustering method.
pbmc <- qc_filter(pbmc,
min_feature = 200,
max_feature = 2500,
max_mito_percent = 5,
min_cell = 3
)
pbmc <- normalize(pbmc, scale_factor = 1e6, log_transformation = TRUE)
pbmc <- find_HVG(pbmc, n_feature = 2000, loess_span = 0.5)
#> Warning in sqrt(predicted_vars): NaNs produced
pbmc <- scale_data(pbmc)
pbmc <- principal_component_analysis(pbmc, PC1 = 3, PC2 = 4)
pbmc <- kmeans(pbmc, k = 7, dimensions = 4)