README.md
In dianalow/nMyo: nMyo
title: "nMyo: an R package and shiny applet to visualise the scRNA-seq expression profiles of mouse cardiac non-myocytes"
author: "Efthymios Motakis & Diana Low"
date: "05/01/2020"
output: html_document
```{r setup, include=FALSE}
library(nMyo)
knitr::opts_chunk$set(echo = TRUE)
# 1. Background
We have developed the **nMyo** R shiny applet and associated R package to visualise the single-cell RNAseq expression profiles of 2031 high-quality mouse cardiac non-myocytes. The quality control, data generation and normalisation, cluster annotation and differential expression steps are briefly discussed in the subsequent paragraphs. **nMyo**'s input is the outcome of the counts normalisation, cell characterisation and differential expression analysis pipelines (the respective input format is described below). Practically, the user selects a gene ID and our software generates a series of interactive plots for data viusalisation and exploratory analysis.
## 1.1 Cell isolation and RNA sequencing
The cells were dissociated from the whole left ventricle of 3 Sham-operated and 3 MI-induced mice. The cell isolation was based on a FACS-scRNAseq protocol for global, non-biased RNA sequencing performed with the SMART-seq2 technology on Illumina 4000.
## 1.2 From fastq to raw counts
The raw single-cell, paired-end reads in fastq format of the originally 2,272 mouse non-myocytes (1,152 MI and 1,120 Sham) were initially processed with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) for quality control at the base and sequence level. The Nextera adapters were removed by `Trimmomatic` (Bolger et al, 2014) and the PCR-deduplication was perfomred by `FastUniq` (Hu et al, 2012). The GENCODE M12 annotated transcripts were quantified by `Kallisto v.43` (Bray et al, 2016). The conversion of transcript to gene counts (and transcript length corrected counts) was performed by the `tximport` package in R.
# 1.3 Overview of the main data analysis
We performed QC at the gene expression level, retaining samples with more than 200,000 reads, less than 15% of the reads mapped to mitochondrial RNA and more than 1500 detected genes. The samples that did not pass the QC were considered of low quality and they were removed from further analysis. The transcript-length corrected gene counts of 26,942 genes and 2,031 high-quality cells (957 Sham and 1,074 MI) were normalized to a set of endogenous reference genes that were identified by intersecting the top non-differentially expressed genes (Risso et al, 2014) and least variable genes of `scatter` (HVG function) in R. To identify the cell types, we run 2D `t-SNE` on the normalised data using the first 20 PC components (Rtsne; van der Maaten, 2014) and clustered the data by `Affinity Propagation` (Frey and Dueck, 2007). The clusters were annotated based on evidence from the significant up-regulation of known markers of major cardiac cell types combined with gene ontology functions of upregulated genes. Eleven distinct cell types were identified, i.e. cardiac fibroblasts (440 Sham and 390 MI), endothelial cells (326 Sham and 258 MI), endothelial ECM cells (61 Sham and 65 MI), lymphatic endothelial (10 Sham and 34 MI), T-cells (16 Sham and 39 MI), B-cells (20 Sham and 17MI), two machrophage populations (Pop1: 4 Sham and 69 MI; Pop2: 17 Sham and 18 MI), smooth muscle cells (12 Sham and 19 MI), pericytes (36 Sham and 46 MI) and neutrophils (5 Sham and 22 MI). A small cluster of 10 Sham and 7 MI cells had an unpredicted cell type. The edgeR differential expression analysis of MI vs Sham or CellType *i* vs Rest raw counts was performed after accounting for technical differences between plates and cell cycle phase via the `cyclone()` function of `scran` R package (FDR = 10%).
# 2. The input matrices
**nMyo** expects as input a set of .txt files of a specific format which we will briefly describe here.
- CorrCounts: a matrix of the normalised gene counts of 26,942 genes and 2,031 cells. The first column (column name: `Marker`) contains the gene IDs in the format *EnsemblID:GeneSymbol*. The column names of the other columns are the unique cell IDs.
- design_table: a matrix summarising the various cell characteristics (experimentally known or predicted) of our dataset. The first column (column name: `SampleID`) contains the unique cell IDs depicted in the corrCounts matrix (in the same order). Some important experimentally known characteristics are the sequencing `Plate` and the cell experimental `Condition` (Sham and MI). The rest have been estimated from the data. The most important are: the cell `Library Size`, the percentage of reads mapped to `mitochondrial` genes, the number of `detected genes`, the estimated `cell cycle` phases, the t-SNE dimensions (`Dim1`, `Dim2` and `Dim3`) and the estimated `cell types`.
- DE: a matrix with the differential expression analysis estimates. The first column (column name: `Marker`) has the gene IDs formatted as in the CorrCounts matrix. Other essential statistics are the `logFC`, `logCPM`, `PValue` and `FDR` that are directly obtained from edgeR. The last two columns, `cell type` and `comparison`, contain values to characterise and separate the pairwise tests. For example, CellType = `ShamMI` and Comparison = `Fibroblast-Rest` indicate the comparison of Fibroblast vs the rest of the cell types across all cells while CellType = `Fibroblast` and Comparison = `Sham-MI` the comparison of Sham vs MI in fibrblasts.
# 3. The **nMyo** R package
The **nMyo** R package can be installed from GitHub. This will automatically install and load all dependencies. The data visualisation is performed in 3 steps, i.e. data loading, gene selection and data visualisation. Here, we briefly describe the functions of interest. More detailed information can be obtained from the package help pages (e.g. `?readData` for the help page of the readData() function).
## 3.1 Data loading
The data loading step is done as:
```{r loading}
data<-readData(Counts_file=system.file("extdata", "CorrCounts.txt", package = "nMyo"),
Design_file=system.file("extdata", "design_table.txt", package = "nMyo"),
DE_file=system.file("extdata", "DE.txt", package = "nMyo"))
# listed components
names(data)
# the CorrCounts
data$Counts[1:5, 1:5]
# the design_table
data$Design[1:5, ]
# the DE
data$DEstats[1:5, ]
# the data stamp (for file storage)
data$Date_Stamp
Among the data components we find the CorrCounts (Counts), design (Design) and differential expression (DEstats) matrices as well as other automatically generated information such as: Annotation with the geneIDs, logFC with the logFC cut-off for preliminary marker filtering based on the differential expression analysis results (default is 0), FDR with the FDR cut-off for preliminary marker filtering based on the differential expression analysis results (default is 1), Output folder specifying the folder that stores the results and Date stamp that assigns unique filenames for storage.
3.2 Gene selection
Next, the user selects the gene for visualisation by its Ensembl ID or its gene symbol or the combination EnsemblID:GeneSymbol. Any of the genes present in the CorrCounts matrix can be selected. Below, we indicate some examples:
```{r query1}
data_Postn <- MarkerQuery(Data = data, marker = "Postn")
Parameter `Data` accepts the outcome of readData() function while parameter `marker` takes the ID of interest. The above example shows the selection of *Postn* gene for further analysis. Alternatively, the user can select the respective Ensembl ID as:
```{r query2}
data_Postn_alt <- MarkerQuery(Data = data, marker = "ENSMUSG00000027750")
or he can use the combination of the two as:
```{r query3}
data_Postn_alt2 <- MarkerQuery(Data = data, marker = "ENSMUSG00000027750:Postn")
To account for small typos in the `marker` input, **nMyo** finds the best match associated to the selected gene, e.g.:
```{r query4}
data_Postn_err <- MarkerQuery(Data = data, marker = "Post")
If the selected gene is not present in the data, nMyo generates an appropiate error and the analysis stops, prompting the user to select another ID:
```{r query5}
data_err <- MarkerQuery(Data = data, marker = "Tp53")
## 3.3 Data visualisation
**nMyo** can generate four types of interative plots for gene expression visualisation and exploration. First, we will see the t-SNE scatterplot depicting the expression levels of the selected gene with a colour gradient (blue-to-red). Mouse-over on the plot reveals the estimated normalised expression level, the experimental condition and the estimated cell type of each cell. By default, double clicking on any of the dots highlights all cells of the same cell type (controlled by the parameter `highlight.by`). Alternative values of `highlight.by` are possible to bring forth other aspects of the data (e.g. `Condition` or any other factor of the design_table). The logical parameters `show.plot` and `save.plot` determine whether the plot will be shown on screen and/or stored as an html (interactive) file. Finally, clicking on the gene ID of the plot legend directs the user to the respective NCBI page of the gene.
```{r tsne}
scatter_Postn <- doScatterDR(Data = data_Postn,
highlight.by = "CellType",
show.plot = TRUE,
save.plot = FALSE)
A split-violin plot for each cell type (default; controlled by the parameter grouping.by) splitted by Condition (default; controlled by the parameter grouping2.by) can be generated as:
```{r violin}
violin_Postn <- doViolin(Data = data_Postn,
grouping.by = "CellType",
grouping2.by = "Condition",
show.plot = TRUE,
save.plot = FALSE)
Other values of `grouping.by` and `grouping2.by` are also possible (essentially any of the factors of the design_table). For example, setting `grouping2.by = NULL` generates a simple violin plot for each cell type. A constraint is currently imposed on `grouping2.by` that can only accept factors with 2 levels (e.g. `Condition` is either Sham or MI).
```{r violin2}
violin_Postn2 <- doViolin(Data = data_Postn,
grouping.by = "CellType",
grouping2.by = NULL,
show.plot = TRUE,
save.plot = FALSE)
The Volcano and MA plots are governed by a similar set of parameters. An interactive volcano plot can be generated as:
```{r volcano}
volcano_Postn <- doVolcano(Data = data_Postn,
filters = "Comparison=='Fibroblast: Sham-MI'",
logFC = 1,
FDR = 0.01,
show.plot = TRUE,
save.plot = FALSE)
Parameter `filters` specifies the pairwise comparison to be shown. There are several alternatives:
```{r pairwise}
# all pairwise comparisons (to be used in filter parameter)
names(table(data$DEstats$Comparison))
The differentially expressed genes (the logFC and FDR parameters control the cutoffs) are highlighted in different color along with the selected gene which is highlighted with a large triangle. Clicking on it, the user is redirected to the accosiated NCBI page. In the same way, one can generate the MA plot:
```{r ma}
ma_Postn <- doMA(Data = data_Postn,
filters = "Comparison=='Fibroblast: Sham-MI'",
logFC = 1,
FDR = 0.01,
show.plot = TRUE,
save.plot = FALSE)
# 4. The **nMyo** R shiny applet
The **nMyo** web applet can be accessed as [ADD INFOMRATION HERE]. The applet offers to the computationally inexperienced users an alternative, easy-to-use way to visualise our non-myocyte dataset at the cost of less flexibility. It only requires a prior installation of R or R studio in their system.
```{r pressure, echo=FALSE, fig.cap="nMyo welcome screen", out.width = '100%'}
knitr::include_graphics("screen1.png")
The three input data tables are automatically loaded upon nMyo's initialization. By clicking on Get Started! the user is redirected to the Quick Analysis tab.
{r pressure, echo=FALSE, fig.cap="nMyo welcome screen", out.width = '100%'}
knitr::include_graphics("screen2.png")
5. References
The right part of the screen allows the user to download the input matrices and see at a glance the package's functionality. As before, the user is enabled to select the gene ID of interest and load it in the system. The Plot Options tab will be subsequently activated enabling the user to select one of the four data visualisation plots.
By default, the t-SNE higlights the cell types (highlight.by = CellType), the violin is done by cell type only (not split) and the MA/Volcano plots highlight the differentially expressed genes at logFC = 1 and FDR = 0.01. Clicking on the by condition checkbox activates the Conditions highlight in the t-SNE and the split by condition in the violin plot. The cell type, logFC and FDR parameters can be adjusted accoring t the user's needs. The user can either see the plot at the bottom of the screen (Show data) and/or save it (Save data) as an htlm interactive file located in the Data/InteractivePlots subfolder of the package.
-
Bolger et al. "Trimmomatic: a flexible trimmer for Illumina sequence data". Bioinformatics 2014; 30:2114-2120.
-
Hu et al. "FastUniq: a fast de novo duplicates removal tool for paired short reads". PLoS One 2012; 7:e52249.
-
Bray et al. "Near-optimal probabilistic RNA-seq quantification". Nature Biotechnology 2016; 34:525-527.
-
van der Maaten. "Accelerating t-SNE using Tree-Based Algorithms". Journal of Machine Learning Research 2014; 15:3221-3245.
-
Frey and Dueck. "Clustering by passing messages between data points". Science 2007; 315:972-976.
dianalow/nMyo documentation built on June 2, 2020, 12:03 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
title: "nMyo: an R package and shiny applet to visualise the scRNA-seq expression profiles of mouse cardiac non-myocytes" author: "Efthymios Motakis & Diana Low" date: "05/01/2020" output: html_document
```{r setup, include=FALSE} library(nMyo) knitr::opts_chunk$set(echo = TRUE)
# 1. Background
We have developed the **nMyo** R shiny applet and associated R package to visualise the single-cell RNAseq expression profiles of 2031 high-quality mouse cardiac non-myocytes. The quality control, data generation and normalisation, cluster annotation and differential expression steps are briefly discussed in the subsequent paragraphs. **nMyo**'s input is the outcome of the counts normalisation, cell characterisation and differential expression analysis pipelines (the respective input format is described below). Practically, the user selects a gene ID and our software generates a series of interactive plots for data viusalisation and exploratory analysis.
## 1.1 Cell isolation and RNA sequencing
The cells were dissociated from the whole left ventricle of 3 Sham-operated and 3 MI-induced mice. The cell isolation was based on a FACS-scRNAseq protocol for global, non-biased RNA sequencing performed with the SMART-seq2 technology on Illumina 4000.
## 1.2 From fastq to raw counts
The raw single-cell, paired-end reads in fastq format of the originally 2,272 mouse non-myocytes (1,152 MI and 1,120 Sham) were initially processed with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) for quality control at the base and sequence level. The Nextera adapters were removed by `Trimmomatic` (Bolger et al, 2014) and the PCR-deduplication was perfomred by `FastUniq` (Hu et al, 2012). The GENCODE M12 annotated transcripts were quantified by `Kallisto v.43` (Bray et al, 2016). The conversion of transcript to gene counts (and transcript length corrected counts) was performed by the `tximport` package in R.
# 1.3 Overview of the main data analysis
We performed QC at the gene expression level, retaining samples with more than 200,000 reads, less than 15% of the reads mapped to mitochondrial RNA and more than 1500 detected genes. The samples that did not pass the QC were considered of low quality and they were removed from further analysis. The transcript-length corrected gene counts of 26,942 genes and 2,031 high-quality cells (957 Sham and 1,074 MI) were normalized to a set of endogenous reference genes that were identified by intersecting the top non-differentially expressed genes (Risso et al, 2014) and least variable genes of `scatter` (HVG function) in R. To identify the cell types, we run 2D `t-SNE` on the normalised data using the first 20 PC components (Rtsne; van der Maaten, 2014) and clustered the data by `Affinity Propagation` (Frey and Dueck, 2007). The clusters were annotated based on evidence from the significant up-regulation of known markers of major cardiac cell types combined with gene ontology functions of upregulated genes. Eleven distinct cell types were identified, i.e. cardiac fibroblasts (440 Sham and 390 MI), endothelial cells (326 Sham and 258 MI), endothelial ECM cells (61 Sham and 65 MI), lymphatic endothelial (10 Sham and 34 MI), T-cells (16 Sham and 39 MI), B-cells (20 Sham and 17MI), two machrophage populations (Pop1: 4 Sham and 69 MI; Pop2: 17 Sham and 18 MI), smooth muscle cells (12 Sham and 19 MI), pericytes (36 Sham and 46 MI) and neutrophils (5 Sham and 22 MI). A small cluster of 10 Sham and 7 MI cells had an unpredicted cell type. The edgeR differential expression analysis of MI vs Sham or CellType *i* vs Rest raw counts was performed after accounting for technical differences between plates and cell cycle phase via the `cyclone()` function of `scran` R package (FDR = 10%).
# 2. The input matrices
**nMyo** expects as input a set of .txt files of a specific format which we will briefly describe here.
- CorrCounts: a matrix of the normalised gene counts of 26,942 genes and 2,031 cells. The first column (column name: `Marker`) contains the gene IDs in the format *EnsemblID:GeneSymbol*. The column names of the other columns are the unique cell IDs.
- design_table: a matrix summarising the various cell characteristics (experimentally known or predicted) of our dataset. The first column (column name: `SampleID`) contains the unique cell IDs depicted in the corrCounts matrix (in the same order). Some important experimentally known characteristics are the sequencing `Plate` and the cell experimental `Condition` (Sham and MI). The rest have been estimated from the data. The most important are: the cell `Library Size`, the percentage of reads mapped to `mitochondrial` genes, the number of `detected genes`, the estimated `cell cycle` phases, the t-SNE dimensions (`Dim1`, `Dim2` and `Dim3`) and the estimated `cell types`.
- DE: a matrix with the differential expression analysis estimates. The first column (column name: `Marker`) has the gene IDs formatted as in the CorrCounts matrix. Other essential statistics are the `logFC`, `logCPM`, `PValue` and `FDR` that are directly obtained from edgeR. The last two columns, `cell type` and `comparison`, contain values to characterise and separate the pairwise tests. For example, CellType = `ShamMI` and Comparison = `Fibroblast-Rest` indicate the comparison of Fibroblast vs the rest of the cell types across all cells while CellType = `Fibroblast` and Comparison = `Sham-MI` the comparison of Sham vs MI in fibrblasts.
# 3. The **nMyo** R package
The **nMyo** R package can be installed from GitHub. This will automatically install and load all dependencies. The data visualisation is performed in 3 steps, i.e. data loading, gene selection and data visualisation. Here, we briefly describe the functions of interest. More detailed information can be obtained from the package help pages (e.g. `?readData` for the help page of the readData() function).
## 3.1 Data loading
The data loading step is done as:
```{r loading}
data<-readData(Counts_file=system.file("extdata", "CorrCounts.txt", package = "nMyo"),
Design_file=system.file("extdata", "design_table.txt", package = "nMyo"),
DE_file=system.file("extdata", "DE.txt", package = "nMyo"))
# listed components
names(data)
# the CorrCounts
data$Counts[1:5, 1:5]
# the design_table
data$Design[1:5, ]
# the DE
data$DEstats[1:5, ]
# the data stamp (for file storage)
data$Date_Stamp
Among the data components we find the CorrCounts (Counts), design (Design) and differential expression (DEstats) matrices as well as other automatically generated information such as: Annotation with the geneIDs, logFC with the logFC cut-off for preliminary marker filtering based on the differential expression analysis results (default is 0), FDR with the FDR cut-off for preliminary marker filtering based on the differential expression analysis results (default is 1), Output folder specifying the folder that stores the results and Date stamp that assigns unique filenames for storage.
3.2 Gene selection
Next, the user selects the gene for visualisation by its Ensembl ID or its gene symbol or the combination EnsemblID:GeneSymbol. Any of the genes present in the CorrCounts matrix can be selected. Below, we indicate some examples:
```{r query1} data_Postn <- MarkerQuery(Data = data, marker = "Postn")
Parameter `Data` accepts the outcome of readData() function while parameter `marker` takes the ID of interest. The above example shows the selection of *Postn* gene for further analysis. Alternatively, the user can select the respective Ensembl ID as:
```{r query2}
data_Postn_alt <- MarkerQuery(Data = data, marker = "ENSMUSG00000027750")
or he can use the combination of the two as:
```{r query3} data_Postn_alt2 <- MarkerQuery(Data = data, marker = "ENSMUSG00000027750:Postn")
To account for small typos in the `marker` input, **nMyo** finds the best match associated to the selected gene, e.g.:
```{r query4}
data_Postn_err <- MarkerQuery(Data = data, marker = "Post")
If the selected gene is not present in the data, nMyo generates an appropiate error and the analysis stops, prompting the user to select another ID:
```{r query5} data_err <- MarkerQuery(Data = data, marker = "Tp53")
## 3.3 Data visualisation
**nMyo** can generate four types of interative plots for gene expression visualisation and exploration. First, we will see the t-SNE scatterplot depicting the expression levels of the selected gene with a colour gradient (blue-to-red). Mouse-over on the plot reveals the estimated normalised expression level, the experimental condition and the estimated cell type of each cell. By default, double clicking on any of the dots highlights all cells of the same cell type (controlled by the parameter `highlight.by`). Alternative values of `highlight.by` are possible to bring forth other aspects of the data (e.g. `Condition` or any other factor of the design_table). The logical parameters `show.plot` and `save.plot` determine whether the plot will be shown on screen and/or stored as an html (interactive) file. Finally, clicking on the gene ID of the plot legend directs the user to the respective NCBI page of the gene.
```{r tsne}
scatter_Postn <- doScatterDR(Data = data_Postn,
highlight.by = "CellType",
show.plot = TRUE,
save.plot = FALSE)
A split-violin plot for each cell type (default; controlled by the parameter grouping.by) splitted by Condition (default; controlled by the parameter grouping2.by) can be generated as:
```{r violin} violin_Postn <- doViolin(Data = data_Postn, grouping.by = "CellType", grouping2.by = "Condition", show.plot = TRUE, save.plot = FALSE)
Other values of `grouping.by` and `grouping2.by` are also possible (essentially any of the factors of the design_table). For example, setting `grouping2.by = NULL` generates a simple violin plot for each cell type. A constraint is currently imposed on `grouping2.by` that can only accept factors with 2 levels (e.g. `Condition` is either Sham or MI).
```{r violin2}
violin_Postn2 <- doViolin(Data = data_Postn,
grouping.by = "CellType",
grouping2.by = NULL,
show.plot = TRUE,
save.plot = FALSE)
The Volcano and MA plots are governed by a similar set of parameters. An interactive volcano plot can be generated as:
```{r volcano} volcano_Postn <- doVolcano(Data = data_Postn, filters = "Comparison=='Fibroblast: Sham-MI'", logFC = 1, FDR = 0.01, show.plot = TRUE, save.plot = FALSE)
Parameter `filters` specifies the pairwise comparison to be shown. There are several alternatives:
```{r pairwise}
# all pairwise comparisons (to be used in filter parameter)
names(table(data$DEstats$Comparison))
The differentially expressed genes (the logFC and FDR parameters control the cutoffs) are highlighted in different color along with the selected gene which is highlighted with a large triangle. Clicking on it, the user is redirected to the accosiated NCBI page. In the same way, one can generate the MA plot:
```{r ma} ma_Postn <- doMA(Data = data_Postn, filters = "Comparison=='Fibroblast: Sham-MI'", logFC = 1, FDR = 0.01, show.plot = TRUE, save.plot = FALSE)
# 4. The **nMyo** R shiny applet
The **nMyo** web applet can be accessed as [ADD INFOMRATION HERE]. The applet offers to the computationally inexperienced users an alternative, easy-to-use way to visualise our non-myocyte dataset at the cost of less flexibility. It only requires a prior installation of R or R studio in their system.
```{r pressure, echo=FALSE, fig.cap="nMyo welcome screen", out.width = '100%'}
knitr::include_graphics("screen1.png")
The three input data tables are automatically loaded upon nMyo's initialization. By clicking on Get Started! the user is redirected to the Quick Analysis tab.
{r pressure, echo=FALSE, fig.cap="nMyo welcome screen", out.width = '100%'}
knitr::include_graphics("screen2.png")
5. References
The right part of the screen allows the user to download the input matrices and see at a glance the package's functionality. As before, the user is enabled to select the gene ID of interest and load it in the system. The Plot Options tab will be subsequently activated enabling the user to select one of the four data visualisation plots.
By default, the t-SNE higlights the cell types (highlight.by = CellType), the violin is done by cell type only (not split) and the MA/Volcano plots highlight the differentially expressed genes at logFC = 1 and FDR = 0.01. Clicking on the by condition checkbox activates the Conditions highlight in the t-SNE and the split by condition in the violin plot. The cell type, logFC and FDR parameters can be adjusted accoring t the user's needs. The user can either see the plot at the bottom of the screen (Show data) and/or save it (Save data) as an htlm interactive file located in the Data/InteractivePlots subfolder of the package.
-
Bolger et al. "Trimmomatic: a flexible trimmer for Illumina sequence data". Bioinformatics 2014; 30:2114-2120.
-
Hu et al. "FastUniq: a fast de novo duplicates removal tool for paired short reads". PLoS One 2012; 7:e52249.
-
Bray et al. "Near-optimal probabilistic RNA-seq quantification". Nature Biotechnology 2016; 34:525-527.
-
van der Maaten. "Accelerating t-SNE using Tree-Based Algorithms". Journal of Machine Learning Research 2014; 15:3221-3245.
-
Frey and Dueck. "Clustering by passing messages between data points". Science 2007; 315:972-976.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.