Contents

1 Installation

if (!require("BiocManager"))
    install.packages("BiocManager")
BiocManager::install("FuseSOM")

2 Introduction

A correlation based multiview self organizing map for the characterization of cell types (FuseSOM) is a tool for unsupervised clustering. FuseSOM is robust and achieves high accuracy by combining a Self Organizing Map architecture and a Multiview integration of correlation based metrics to cluster highly multiplexed in situ imaging cytometry assays. The FuseSOM pipeline has been streamlined and accepts currently used data structures including SingleCellExperiment and SpatialExperiment objects as well as DataFrames.

3 Disclaimer

This is purely a tool generated for clustering and as such it does not provide any means for QC and feature selection. It is advisable that the user first use other tools for quality control and feature selection before running FuseSOM.

4 Getting Started

4.1 FuseSOM Matrix Input

If you have a matrix containing expression data that was QCed and normalised by some other tool, the next step is to run the FuseSOM algorithm.This can be done by calling the runFuseSOM() function which takes in the matrix of interest where the columns are markers and the rows are observations, the makers of interest (if this is not provided, it is assumed that all columns are markers), and the number of clusters.

# load FuseSOM
library(FuseSOM)

Next we will load in the Risom et al dataset and run it through the FuseSOM pipeline. This dataset profiles the spatial landscape of ductal carcinoma in situ (DCIS), which is a pre-invasive lesion that is thought to be a precursor to invasive breast cancer (IBC). The key conclusion of this manuscript (amongst others) is that spatial information about cells can be used to predict disease progression in patients.We will also be using the markers used in the original study.

# load in the data
data("risom_dat")

# define the markers of interest
risomMarkers <- c('CD45','SMA','CK7','CK5','VIM','CD31','PanKRT','ECAD',
                   'Tryptase','MPO','CD20','CD3','CD8','CD4','CD14','CD68','FAP',
                   'CD36','CD11c','HLADRDPDQ','P63','CD44')

# we will be using the manual_gating_phenotype as the true cell type to gauge 
# performance
names(risom_dat)[names(risom_dat) == 'manual_gating_phenotype'] <- 'CellType'

Now that we have loaded the data and define the markers of interest. We can run the FuseSOM algorithm. We have provided a function runFuseSOM that runs the pipeline from top to bottom and returns the cluster labels as well as the Self Organizing Map model.

risomRes <- runFuseSOM(data = risom_dat, markers = risomMarkers, 
                        numClusters = 23)
## You have provided a dataset of class data.frame
## Everything looks good. Now running the FuseSOM algorithm
## Now Generating the Self Organizing Map Grid
## Optimal Grid Size is: 8
## Now Running the Self Organizing Map Model
## Now Clustering the Prototypes
## Loading required namespace: fastcluster
## Now Mapping Clusters to the Original Data
## The Prototypes have been Clustered and Mapped Successfully
## The FuseSOM algorithm has completed successfully

Lets look at the distribution of the clusters.

# get the distribution of the clusters
table(risomRes$clusters)/sum(table(risomRes$clusters))
## 
##   cluster_1  cluster_10  cluster_11  cluster_12  cluster_13  cluster_14 
## 0.324491905 0.012601906 0.021170628 0.017094385 0.014539557 0.015472500 
##  cluster_15  cluster_16  cluster_17  cluster_18  cluster_19   cluster_2 
## 0.014625675 0.010922609 0.016491560 0.005741187 0.046460558 0.071506488 
##  cluster_20  cluster_21  cluster_22  cluster_23   cluster_3   cluster_4 
## 0.042197726 0.020165920 0.013908026 0.005138363 0.053393042 0.064746240 
##   cluster_5   cluster_6   cluster_7   cluster_8   cluster_9 
## 0.085055690 0.027141463 0.048685268 0.061430704 0.007018601

Looks like cluster_1 has about \(32\%\) of the cells which is interesting. Next, lets generate a heatmap of the marker expression for each cluster.

risomHeat <- FuseSOM::markerHeatmap(data = risom_dat, markers = risomMarkers,
                            clusters = risomRes$clusters, clusterMarkers = TRUE)

4.2 Using FuseSOM to estimate the number of clusters

FuseSOM also provides functionality for estimating the number of clusters in a dataset using three classes of methods including:

  1. Discriminant based method.
    • A method developed in house based on discriminant based maximum clusterability projection pursuit
  2. Distance based methods which includes:
    • The Gap Statistic
    • The Jump Statistic
    • The Slope Statistic
    • The Within Cluster Dissimilarity Statistic
    • The Silhouette Statistic

We can estimate the number of clusters using the estimateNumCluster. Run help(estimateNumCluster) to see it’s complete functionality.

# lets estimate the number of clusters using all the methods
# original clustering has 23 clusters so we will set kseq from 2:25
# we pass it the som model generated in the previous step
risomKest <- estimateNumCluster(data = risomRes$model, kSeq = 2:25, 
                                  method = c("Discriminant", "Distance"))
## Now Computing the Number of Clusters using Discriminant Analysis
## Now Computing The Number Of Clusters Using Distance Analysis

We can then use this result to determine the best number of clusters for this dataset based on the different metrics. The FuseSOM package provides a plotting function (optiPlot) which generates an elbow plot with the optimal value for the number of clusters for the distance based methods. See below

# what is the best number of clusters determined by the discriminant method?
# optimal number of clusters according to the discriminant method is 7
risomKest$Discriminant 
## [1] 11
# we can plot the results using the optiplot function
pSlope <- optiPlot(risomKest, method = 'slope')
pSlope

pJump <- optiPlot(risomKest, method = 'jump')
pJump

pWcd <- optiPlot(risomKest, method = 'wcd')
pWcd

pGap <- optiPlot(risomKest, method = 'gap')
pGap

pSil <- optiPlot(risomKest, method = 'silhouette')
pSil

From the plots, we see that the Jump statistics almost perfectly capture the number of clusters. The Gap method is a close second with \(15\) clusters. All the other methods significantly underestimates the number of clusters.

4.3 FuseSOM Sinlge Cell Epxeriment object as input.

The FuseSOM algorithm is also equipped to take in a SingleCellExperiment object as input. The results of the pipeline will be written to either the metada or the colData fields. See below.

First we create a SingleCellExperiment object

library(SingleCellExperiment)

# create a singelcellexperiment object
colDat <- risom_dat[, setdiff(colnames(risom_dat), risomMarkers)]
sce <- SingleCellExperiment(assays = list(counts = t(risom_dat)),
                                 colData = colDat)

sce
## class: SingleCellExperiment 
## dim: 23 69672 
## metadata(0):
## assays(1): counts
## rownames(23): CD45 SMA ... CD44 CellType
## rowData names(0):
## colnames: NULL
## colData names(1): X
## reducedDimNames(0):
## mainExpName: NULL
## altExpNames(0):

Next we pass it to the runFuseSOM() function. Here, we can provide the assay in which the data is stored and what name to store the clusters under in the colData section. Note that the Self Organizing Map that is generated will be stored in the metadata field.

risomRessce <- runFuseSOM(sce, markers = risomMarkers, assay = 'counts', 
                      numClusters = 23, verbose = FALSE)
## You have provided a dataset of class SingleCellExperiment
## Everything looks good. Now running the FuseSOM algorithm
## Now Generating the Self Organizing Map Grid
## Optimal Grid Size is: 8
## Now Running the Self Organizing Map Model
## Now Clustering the Prototypes
## Now Mapping Clusters to the Original Data
## The Prototypes have been Clustered and Mapped Successfully
## The FuseSOM algorithm has completed successfully
colnames(colData(risomRessce))
## [1] "X"        "clusters"
names(metadata(risomRessce))
## [1] "SOM"

Notice how the there is now a clusters column in the colData and SOM field in the metadata. You can run this function again with a new set of cluster number. If you provide a new name for the clusters, it will be stored under that new column, else, it will overwrite the current clusters column. Running it again on the same object will overwrite the SOM field in the metadata.

Just like before, lets plot the heatmap of the resulting clusters across all markers.

data <- risom_dat[, risomMarkers] # get the original data used
clusters <- colData(risomRessce)$clusters # extract the clusters from the sce object
# generate the heatmap
risomHeatsce <- markerHeatmap(data = risom_dat, markers = risomMarkers,
                            clusters = clusters, clusterMarkers = TRUE)

4.4 Using FuseSOM to estimate the number of clusters for single cell experiment objects

Just like before, we can estimate the number of clusters

# lets estimate the number of clusters using all the methods
# original clustering has 23 clusters so we will set kseq from 2:25
# now we pass it a singlecellexperiment object instead of the som model as before
# this will return a singelcellexperiment object where the metatdata contains the
# cluster estimation information
risomRessce <- estimateNumCluster(data = risomRessce, kSeq = 2:25, 
                                  method = c("Discriminant", "Distance"))
## You have provided a dataset of class: SingleCellExperiment
## Now Computing the Number of Clusters using Discriminant Analysis
## Now Computing The Number Of Clusters Using Distance Analysis
names(metadata(risomRessce))
## [1] "SOM"               "clusterEstimation"

Notice how the metadata now contains a clusterEstimation field which holds the results from the estimateNumCluster() function

We can assess the results in a similar fashion as before

# what is the best number of clusters determined by the discriminant method?
# optimal number of clusters according to the discriminant method is 8
metadata(risomRessce)$clusterEstimation$Discriminant 
## [1] 11
# we can plot the results using the optiplot function
pSlope <- optiPlot(risomRessce, method = 'slope')
## You have provided a dataset of class: SingleCellExperiment
pSlope

pJump <- optiPlot(risomRessce, method = 'jump')
## You have provided a dataset of class: SingleCellExperiment
pJump

pWcd <- optiPlot(risomRessce, method = 'wcd')
## You have provided a dataset of class: SingleCellExperiment
pWcd

pGap <- optiPlot(risomRessce, method = 'gap')
## You have provided a dataset of class: SingleCellExperiment
pGap

pSil <- optiPlot(risomRessce, method = 'silhouette')
## You have provided a dataset of class: SingleCellExperiment
pSil

Again, we see that the Jump statistics almost perfectly capture the number of clusters. The Gap method is a close second with \(15\) clusters. All the other methods significantly underestimates the number of clusters.

4.5 FuseSOM Spatial Epxeriment object as input.

The methodology for Spatial Epxeriment is exactly the same as that of Single Cell Epxeriment

5 sessionInfo()

sessionInfo()
## R version 4.2.2 (2022-10-31)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.5 LTS
## 
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.16-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.16-bioc/R/lib/libRlapack.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_GB              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] stats4    stats     graphics  grDevices utils     datasets  methods  
## [8] base     
## 
## other attached packages:
##  [1] SingleCellExperiment_1.20.0 SummarizedExperiment_1.28.0
##  [3] Biobase_2.58.0              GenomicRanges_1.50.1       
##  [5] GenomeInfoDb_1.34.3         IRanges_2.32.0             
##  [7] S4Vectors_0.36.0            BiocGenerics_0.44.0        
##  [9] MatrixGenerics_1.10.0       matrixStats_0.63.0         
## [11] FuseSOM_1.0.1               knitr_1.41                 
## [13] BiocStyle_2.26.0           
## 
## loaded via a namespace (and not attached):
##   [1] colorspace_2.0-3         ggsignif_0.6.4           class_7.3-20            
##   [4] modeltools_0.2-23        mclust_6.0.0             XVector_0.38.0          
##   [7] proxy_0.4-27             ggpubr_0.5.0             farver_2.1.1            
##  [10] flexmix_2.3-18           fansi_1.0.3              splines_4.2.2           
##  [13] mnormt_2.1.1             cachem_1.0.6             brglm_0.7.2             
##  [16] robustbase_0.95-0        jsonlite_1.8.3           broom_1.0.1             
##  [19] cluster_2.1.4            kernlab_0.9-31           pheatmap_1.0.12         
##  [22] BiocManager_1.30.19      compiler_4.2.2           backports_1.4.1         
##  [25] assertthat_0.2.1         Matrix_1.5-3             fastmap_1.1.0           
##  [28] profileModel_0.6.1       cli_3.4.1                htmltools_0.5.3         
##  [31] tools_4.2.2              gtable_0.3.1             glue_1.6.2              
##  [34] GenomeInfoDbData_1.2.9   dplyr_1.0.10             FCPS_1.3.1              
##  [37] DataVisualizations_1.2.3 Rcpp_1.0.9               carData_3.0-5           
##  [40] jquerylib_0.1.4          vctrs_0.5.1              nlme_3.1-160            
##  [43] fpc_2.2-9                psych_2.2.9              xfun_0.35               
##  [46] fastcluster_1.2.3        stringr_1.4.1            lifecycle_1.0.3         
##  [49] rstatix_0.7.1            princurve_2.1.6          DEoptimR_1.0-11         
##  [52] zlibbioc_1.44.0          MASS_7.3-58.1            scales_1.2.1            
##  [55] parallel_4.2.2           RColorBrewer_1.1-3       yaml_2.3.6              
##  [58] ggplot2_3.4.0            yulab.utils_0.0.5        sass_0.4.3              
##  [61] stringi_1.7.8            highr_0.9                permute_0.9-7           
##  [64] rlang_1.0.6              pkgconfig_2.0.3          prabclus_2.3-2          
##  [67] bitops_1.0-7             evaluate_0.18            lattice_0.20-45         
##  [70] purrr_0.3.5              labeling_0.4.2           tidyselect_1.2.0        
##  [73] magrittr_2.0.3           bookdown_0.30            R6_2.5.1                
##  [76] magick_2.7.3             generics_0.1.3           DelayedArray_0.24.0     
##  [79] DBI_1.1.3                pillar_1.8.1             withr_2.5.0             
##  [82] mgcv_1.8-41              abind_1.4-5              RCurl_1.98-1.9          
##  [85] sp_1.5-1                 nnet_7.3-18              tibble_3.1.8            
##  [88] analogue_0.17-6          car_3.1-1                utf8_1.2.2              
##  [91] rmarkdown_2.18           grid_4.2.2               vegan_2.6-4             
##  [94] digest_0.6.30            diptest_0.76-0           tidyr_1.2.1             
##  [97] gridGraphics_0.5-1       coop_0.6-3               munsell_0.5.0           
## [100] ggplotify_0.1.0          bslib_0.4.1