Single-cell sequencing provides novel means to interpret the transcriptomic profiles of individual cells. To obtain in-depth analysis of single-cell sequencing, it requires effective computational methods to accurately predict single-cell clusters because single-cell sequencing techniques only provide the transcriptomic profiles of each cell. Although the accurate estimation of the cell-to-cell similarity is an essential first step to derive reliable single-cell clustering results, it is challenging to obtain accurate similarity measurements because it highly depends on the selection of genes for similarity evaluations and the optimal set of genes for the accurate similarity estimation is typically unknown. Moreover, due to technical limitations, single-cell sequencing includes a larger number of artificial zeros and the technical noise makes it difficult to develop effective single-cell clustering algorithms. Here, we describe a novel single-cell clustering algorithm that can accurately predict single-cell clusters in large-scale single-cell sequencing by effectively reducing the zero-inflated noise and accurately estimating the cell-to-cell similarities. First, we construct an ensemble similarity network based on different similarity estimates, and reduce the artificial noise using a random walk with restart approach. Finally, starting from a larger number small but highly consistent clusters, we iteratively merge a pair of clusters with the maximum similarities until it reaches the predicted number of clusters. Extensive performance evaluation shows that the proposed single-cell clustering algorithm can yield the accurate single-cell clustering results and it can help deciphering the key messages underlying complex biological mechanisms.
Fig. 1 Graphical overview of SICLEN
To install SICLEN (R package), you need to install devtools and type the following command:
install.packages("devtools")
library(devtools)
install_github("jeonglab/SICLEN")
To test the package, we will use the sample single-cell sequencing (Usoskin et al. 2015) and verify the clustering results through low-dimensional visualization
To begin with, we need to load the R package SICLEN and sample single-cell RNA sequencing data. Please type the following code:
library(SICLEN)
data(usoskin)
The sample data includes two arguments: i) gene expression counts and ii) true cell type labels.
> usoskin$counts
L128_D01 L128_H01 L128_B02 L128_C02 L128_E02 L128_A03 L128_B03 L128_D03 L128_E03 L128_C04
Xkr4 0.00 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
L128_F04 L128_G04 L128_B05 L128_D05 L128_G05 L128_D06 L128_F06 L128_H07 L128_A08 L128_H08
Xkr4 0.000 0.000 0.000 0.000 32.027 0.000 0.000 0.000 0.000 0.000
[ reached getOption("max.print") -- omitted 19533 rows ]
> usoskin$label
[1] "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF"
[20] "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF" "NF"
....
[609] "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH" "TH"
Next, run SICLEN to obtain the accurate single-cell clustering resutls. We only need to type the following command.
clustering <- siclen(usoskin$counts)
Once the proposed single-cell clustering algorithm finished, it returns the following objects:
clustering$counts # noise reduced gene expression counts (cpm normalized)
clustering$network # learned ensemble similarity network
clustering$clusterid # predicted clustering labels
clustering$fgenes # potential feature genes that are employed to construc the ensemble similarity network
In order to verify the clustering results, please take a look at clustering$clusterid. It provides the numeric ID for derived single-cell clustering results.
Finally, to verify the clustering results, we can obtain the low-dimensional visualize of single-cell clustering results through t-SNE. In this example, we will use the R packages ggplot2 and Rtnse. Please refer to the following sample code:
# load required packages
library(ggplot2)
library(Rtsne)
# run t-SNE
tsne <- Rtsne(t(log2(1+clustering$counts[clustering$fgenes, ])))
df.tsne <- data.frame(tsne$Y, label = factor(clustering$clusterid))
# visualization using t-SNE
ggplot(df.tsne,aes(x = X1, y = X2, col = factor(label)))+
geom_point(size = 1, alpha = 0.7)+
labs(x= "Dim_1", y = "Dim_2", title = 'Low-dim. visualization (t-SNE)')+
labs(colour = "Label")
Fig. 2 Low-dimensional visualization of the clustering resutls
Here, we have the default parameter setting for the proposed single-cell clustering algorithm. We carefully select the default parameters based on our experimental results. However, different parameters can yield an improved single-cell clustering results depending on the various factors. Please feel free to adjust the parameters.
Usoskin, Dmitry, et al. "Unbiased classification of sensory neuron types by large-scale single-cell RNA sequencing." Nature Neuroscience 18.1 (2015): 145-153.
Hyundoo Jeong, Sungtae Shin, and Hong Gi Yeom "Accurate single-cell clustering through effective noise reductionover ensemble similarity network" Submitted to review.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.