Package 'SC3'

Description

Cell type annotations for data extracted from a publication by Yan et al.

Usage

ann

Format

An object of class data.frame with 90 rows and 1 columns.

Source

http://dx.doi.org/10.1038/nsmb.2660

Each row corresponds to a single cell from 'yan' dataset

Description

Distance between the cells, i.e. columns, in the input expression matrix are calculated using the Euclidean, Pearson and Spearman metrics to construct distance matrices.

Usage

calculate_distance(data, method)

Arguments

Value

distance matrix

Description

Stability index shows how stable each cluster is accross the selected range of k. The stability index varies between 0 and 1, where 1 means that the same cluster appears in every solution for different k.

Usage

calculate_stability(consensus, k)

Arguments

Details

Imagine a given cluster is split into N clusters when k is changed (all possible values of k are provided via ks argument in the main sc3 function). In each of the new clusters there are given_cells of the given cluster and also some extra_cells from other clusters. Then we define stability as follows:

$\frac{1}{ks*N^2}\sum_{ks}\sum_{N}\frac{given\_cells}{given\_cells + extra\_cells}$

Where one N corrects for the number of clusters and the other N is a penalty for splitting the cluster. ks corrects for the range of k.

Value

a numeric vector containing a stability index of each cluster

Description

Consensus matrix is calculated using the Cluster-based Similarity Partitioning Algorithm (CSPA). For each clustering solution a binary similarity matrix is constructed from the corresponding cell labels: if two cells belong to the same cluster, their similarity is 1, otherwise the similarity is 0. A consensus matrix is calculated by averaging all similarity matrices.

Usage

consensus_matrix(clusts)

Arguments

Value

consensus matrix

Description

Computes consensus matrix given cluster labels

Usage

consmx(dat)

Arguments

Description

Used in consmx function

Usage

ED1(x)

Arguments

Description

Used in sc3-funcs.R distance matrix calculation and within the consensus clustering.

Usage

ED2(x)

Arguments

Description

The function finds the eigenvalues of the sample covariance matrix. It will then return the number of significant eigenvalues according to the Tracy-Widom test.

Usage

estkTW(dataset)

Arguments

Value

an estimated number of clusters k

Description

For a given gene a binary classifier is constructed based on the mean cluster expression values (these are calculated using the cell labels). The classifier prediction is then calculated using the gene expression ranks. The area under the receiver operating characteristic (ROC) curve is used to quantify the accuracy of the prediction. A p-value is assigned to each gene by using the Wilcoxon signed rank test.

Usage

get_auroc(gene, labels)

Arguments

Description

Wrapper for calculating biological properties

Usage

get_biolgy(dataset, labels, regime)

Arguments

Value

results of either

Description

Differential expression is calculated using the non-parametric Kruskal-Wallis test. A significant p-value indicates that gene expression in at least one cluster stochastically dominates one other cluster. Note that the calculation of differential expression after clustering can introduce a bias in the distribution of p-values, and thus we advise to use the p-values for ranking the genes only.

Usage

get_de_genes(dataset, labels)

Arguments

Value

a numeric vector containing the differentially expressed genes and correspoding p-values

Examples

d <- get_de_genes(yan[1:10,], as.numeric(ann[,1]))
head(d)

Description

Find marker genes in the dataset. The get_auroc is used to calculate marker values for each gene.

Usage

get_marker_genes(dataset, labels)

Arguments

Value

data.frame containing the marker genes, corresponding cluster indexes and adjusted p-values

Examples

d <- get_marker_genes(yan[1:10,], as.numeric(ann[,1]))
d

Description

Outlier cells in each cluster are detected using robust distances, calculated using the minimum covariance determinant (MCD), namely using covMcd. The outlier score shows how different a cell is from all other cells in the cluster and it is defined as the differences between the square root of the robust distance and the square root of the 99.99

Usage

get_outl_cells(dataset, labels)

Arguments

Value

a numeric vector containing the cell labels and correspoding outlier scores ordered by the labels

Examples

d <- get_outl_cells(yan[1:10,], as.numeric(ann[,1]))
head(d)

Description

Takes data from the logcounts slot, removes spike-ins and applies the gene filter.

Usage

get_processed_dataset(object)

Arguments

Description

Reorders the rows of the input data.frame based on the sc3_k_markers_clusts column and also keeps only the top 10 genes for each value of sc3_k_markers_clusts.

Usage

markers_for_heatmap(markers)

Arguments

Description

Calculate graph Laplacian of a symmetrix matrix

Usage

norm_laplacian(A)

Arguments

Description

This functions returns all marker gene columns from the phenoData slot of the input object corresponding to the number of clusters k. Additionally, it rearranges genes by the cluster index and order them by the area under the ROC curve value inside of each cluster.

Usage

organise_de_genes(object, k, p_val)

Arguments

Description

This functions returns all marker gene columns from the phenoData slot of the input object corresponding to the number of clusters k. Additionally, it rearranges genes by the cluster index and order them by the area under the ROC curve value inside of each cluster.

Usage

organise_marker_genes(object, k, p_val, auroc)

Arguments

Description

Defines train and study cell indeces based on the svm_num_cells and svm_train_inds input parameters

Usage

prepare_for_svm(N, svm_num_cells = NULL, svm_train_inds = NULL, svm_max)

Arguments

Value

A list of indeces of the train and the study cells

Description

Given an hclust object and the number of clusters k this function reindex the clusters inferred by cutree(hc, k)[hc$order], so that they appear in ascending order. This is particularly useful when plotting heatmaps in which the clusters should be numbered from left to right.

Usage

reindex_clusters(hc, k)

Arguments

Examples

hc <- hclust(dist(USArrests), 'ave')
cutree(hc, 10)[hc$order]
reindex_clusters(hc, 10)[hc$order]

Description

This function is a wrapper that executes all steps of SC3 analysis in one go.

Usage

sc3.SingleCellExperiment(object, ks, gene_filter, pct_dropout_min,
  pct_dropout_max, d_region_min, d_region_max, svm_num_cells, svm_train_inds,
  svm_max, n_cores, kmeans_nstart, kmeans_iter_max, k_estimator, biology,
  rand_seed)

## S4 method for signature 'SingleCellExperiment'
sc3(object, ks = NULL, gene_filter = TRUE,
  pct_dropout_min = 10, pct_dropout_max = 90, d_region_min = 0.04,
  d_region_max = 0.07, svm_num_cells = NULL, svm_train_inds = NULL,
  svm_max = 5000, n_cores = NULL, kmeans_nstart = NULL,
  kmeans_iter_max = 1e+09, k_estimator = FALSE, biology = FALSE,
  rand_seed = 1)

Arguments

Value

an object of SingleCellExperiment class

Description

This function calculates differentially expressed (DE) genes, marker genes and cell outliers based on the consensus SC3 clusterings.

Usage

sc3_calc_biology.SingleCellExperiment(object, ks, regime)

## S4 method for signature 'SingleCellExperiment'
sc3_calc_biology(object, ks = NULL,
  regime = NULL)

Arguments

Details

DE genes are calculated using get_de_genes. Results of the DE analysis are saved as new columns in the featureData slot of the input object. The column names correspond to the adjusted p-values of the genes and have the following format: sc3_k_de_padj, where k is the number of clusters.

Marker genes are calculated using get_marker_genes. Results of the marker gene analysis are saved as three new columns (for each k) to the featureData slot of the input object. The column names correspond to the SC3 cluster labels, to the adjusted p-values of the genes and to the area under the ROC curve and have the following format: sc3_k_markers_clusts, sc3_k_markers_padj and sc3_k_markers_auroc, where k is the number of clusters.

Outlier cells are calculated using get_outl_cells. Results of the cell outlier analysis are saved as new columns in the phenoData slot of the input object. The column names correspond to the log2(outlier_score) and have the following format: sc3_k_log2_outlier_score, where k is the number of clusters.

Additionally, biology item is added to the sc3 slot and is set to TRUE indicating that the biological analysis of the dataset has been performed.

Value

an object of SingleCellExperiment class

Description

This function calculates consensus matrices based on the clustering solutions contained in the kmeans item of the sc3 slot of the metadata(object). It then creates and populates the consensus item of the sc3 slot with consensus matrices, their hierarchical clusterings in hclust objects, and Silhouette indeces of the clusters. It also removes the previously calculated kmeans clusterings from the sc3 slot, as they are not needed for further analysis.

Usage

sc3_calc_consens.SingleCellExperiment(object)

## S4 method for signature 'SingleCellExperiment'
sc3_calc_consens(object)

Arguments

Details

Additionally, it also adds new columns to the colData slot of the input object. The column names correspond to the consensus cell labels and have the following format: sc3_k_clusters, where k is the number of clusters.

Value

an object of SingleCellExperiment class

Description

This function calculates distances between the cells. It creates and populates the following items of the sc3 slot of the metadata(object):

• distances - contains a list of distance matrices corresponding to Euclidean, Pearson and Spearman distances.

Usage

sc3_calc_dists.SingleCellExperiment(object)

## S4 method for signature 'SingleCellExperiment'
sc3_calc_dists(object)

Arguments

Value

an object of SingleCellExperiment class

Description

This function transforms all distances items of the sc3 slot of the metadata(object) using either principal component analysis (PCA) or by calculating the eigenvectors of the associated graph Laplacian. The columns of the resulting matrices are then sorted in descending order by their corresponding eigenvalues. The first d columns (where d = max(metadata(object)$sc3$n_dim)) of each transformation are then written to the transformations item of the sc3 slot. Additionally, this function also removes the previously calculated distances from the sc3 slot, as they are not needed for further analysis.

Usage

sc3_calc_transfs.SingleCellExperiment(object)

## S4 method for signature 'SingleCellExperiment'
sc3_calc_transfs(object)

Arguments

Value

an object of SingleCellExperiment class

Description

Uses Tracy-Widom theory on random matrices to estimate the optimal number of clusters k. It creates and populates the k_estimation item of the sc3 slot of the metadata(object).

Usage

sc3_estimate_k.SingleCellExperiment(object)

## S4 method for signature 'SingleCellExperiment'
sc3_estimate_k(object)

Arguments

Value

an estimated value of k

Description

This function writes all SC3 results to an excel file.

Usage

sc3_export_results_xls.SingleCellExperiment(object, filename)

## S4 method for signature 'SingleCellExperiment'
sc3_export_results_xls(object,
  filename = "sc3_results.xls")

Arguments

Description

Runs interactive shiny session of SC3 based on precomputed clusterings.

Usage

sc3_interactive.SingleCellExperiment(object)

## S4 method for signature 'SingleCellExperiment'
sc3_interactive(object)

Arguments

Value

Opens a browser window with an interactive shiny app and visualize all precomputed clusterings.

Description

This function performs kmeans clustering of the matrices contained in the transformations item of the sc3 slot of the metadata(object). It then creates and populates the following items of the sc3 slot:

• kmeans - contains a list of kmeans clusterings.

Usage

sc3_kmeans.SingleCellExperiment(object, ks)

## S4 method for signature 'SingleCellExperiment'
sc3_kmeans(object, ks = NULL)

Arguments

Value

an object of SingleCellExperiment class

Description

Stability index shows how stable each cluster is accross the selected range of ks. The stability index varies between 0 and 1, where 1 means that the same cluster appears in every solution for different k.

Usage

sc3_plot_cluster_stability.SingleCellExperiment(object, k)

## S4 method for signature 'SingleCellExperiment'
sc3_plot_cluster_stability(object, k)

Arguments

Description

The consensus matrix is a NxN matrix, where N is the number of cells. It represents similarity between the cells based on the averaging of clustering results from all combinations of clustering parameters. Similarity 0 (blue) means that the two cells are always assigned to different clusters. In contrast, similarity 1 (red) means that the two cells are always assigned to the same cluster. The consensus matrix is clustered by hierarchical clustering and has a diagonal-block structure. Intuitively, the perfect clustering is achieved when all diagonal blocks are completely red and all off-diagonal elements are completely blue.

Usage

sc3_plot_consensus.SingleCellExperiment(object, k, show_pdata)

## S4 method for signature 'SingleCellExperiment'
sc3_plot_consensus(object, k,
  show_pdata = NULL)

Arguments

Description

SC3 plots gene expression profiles of the 50 genes with the lowest p-values.

Usage

sc3_plot_de_genes.SingleCellExperiment(object, k, p.val, show_pdata)

## S4 method for signature 'SingleCellExperiment'
sc3_plot_de_genes(object, k, p.val = 0.01,
  show_pdata = NULL)

Arguments

Description

The expression panel represents the original input expression matrix (cells in columns and genes in rows) after the gene filter. Genes are clustered by kmeans with k = 100 (dendrogram on the left) and the heatmap represents the expression levels of the gene cluster centers after log2-scaling.

Usage

sc3_plot_expression.SingleCellExperiment(object, k, show_pdata)

## S4 method for signature 'SingleCellExperiment'
sc3_plot_expression(object, k,
  show_pdata = NULL)

Arguments

Description

By default the genes with the area under the ROC curve (AUROC) > 0.85 and with the p-value < 0.01 are selected and the top 10 marker genes of each cluster are visualized in this heatmap.

Usage

sc3_plot_markers.SingleCellExperiment(object, k, auroc, p.val, show_pdata)

## S4 method for signature 'SingleCellExperiment'
sc3_plot_markers(object, k, auroc = 0.85,
  p.val = 0.01, show_pdata = NULL)

Arguments

Description

A silhouette is a quantitative measure of the diagonality of the consensus matrix. An average silhouette width (shown at the bottom left of the silhouette plot) varies from 0 to 1, where 1 represents a perfectly block-diagonal consensus matrix and 0 represents a situation where there is no block-diagonal structure. The best clustering is achieved when the average silhouette width is close to 1.

Usage

sc3_plot_silhouette.SingleCellExperiment(object, k)

## S4 method for signature 'SingleCellExperiment'
sc3_plot_silhouette(object, k)

Arguments

Description

This function prepares an object of SingleCellExperiment class for SC3 clustering. It creates and populates the following items of the sc3 slot of the metadata(object):

• kmeans_iter_max - the same as the kmeans_iter_max argument.

• kmeans_nstart - the same as the kmeans_nstart argument.

• n_dim - contains numbers of the number of eigenvectors to be used in kmeans clustering.

• rand_seed - the same as the rand_seed argument.

• svm_train_inds - if SVM is used this item contains indexes of the training cells to be used for SC3 clustering and further SVM prediction.

• svm_study_inds - if SVM is used this item contains indexes of the cells to be predicted by SVM.

• n_cores - the same as the n_cores argument.

Usage

sc3_prepare.SingleCellExperiment(object, gene_filter, pct_dropout_min,
  pct_dropout_max, d_region_min, d_region_max, svm_num_cells, svm_train_inds,
  svm_max, n_cores, kmeans_nstart, kmeans_iter_max, rand_seed)

## S4 method for signature 'SingleCellExperiment'
sc3_prepare(object, gene_filter = TRUE,
  pct_dropout_min = 10, pct_dropout_max = 90, d_region_min = 0.04,
  d_region_max = 0.07, svm_num_cells = NULL, svm_train_inds = NULL,
  svm_max = 5000, n_cores = NULL, kmeans_nstart = NULL,
  kmeans_iter_max = 1e+09, rand_seed = 1)

Arguments

Value

an object of SingleCellExperiment class

Description

This method parallelize SVM prediction for each k (the number of clusters). Namely, for each k, support_vector_machines function is utilized to predict the labels of study cells. Training cells are selected using svm_train_inds item of the sc3 slot of the metadata(object).

Usage

sc3_run_svm.SingleCellExperiment(object, ks)

## S4 method for signature 'SingleCellExperiment'
sc3_run_svm(object, ks = NULL)

Arguments

Details

Results are written to the sc3_k_clusters columns to the colData slot of the input object, where k is the number of clusters.

Value

an object of SingleCellExperiment class

Description

Train an SVM classifier on a training dataset (train) and then classify a study dataset (study) using the classifier.

Usage

support_vector_machines(train, study, kern)

Arguments

Value

classification of the study dataset

Description

Given matrix A, the procedure returns A'A.

Usage

tmult(x)

Arguments

Description

All distance matrices are transformed using either principal component analysis (PCA) or by calculating the eigenvectors of the graph Laplacian (Spectral). The columns of the resulting matrices are then sorted in descending order by their corresponding eigenvalues.

Usage

transformation(dists, method)

Arguments

Value

transformed distance matrix

Description

Single cell RNA-Seq data extracted from a publication by Yan et al.

Usage

yan

Format

An object of class data.frame with 20214 rows and 90 columns.

Source

http://dx.doi.org/10.1038/nsmb.2660

Columns represent cells, rows represent genes expression values.