Skip to contents

Ordering and Clustering

Source: vignettes/ordering-and-clustering.Rmd
ordering-and-clustering.Rmd

This vignette describes options for ordering rows and columns of an NG-CHM, and the additional options available when the ordering is hierarchical clustering. The YouTube video How to Use Any Label Order or Clustering Method with NG-CHMs, also covers these topics.

These examples build on the setup from Getting Started.

Specifying Row / Column Order

The function chmNew() has optional arguments “rowOrder” and “colOrder” to specify the order of the rows and columns, respectively. Values for these arguments can be a vector, dendrogram, or function specifying the order of the rows/columns.

For example, the following code block creates an NG-CHM with rows sorted alphabetically and columns ordered randomly:

rowOrder <- sort(rownames(matrix_data))
colOrder <- colnames(matrix_data)[sample.int(length(colnames(matrix_data)))]
hm <- chmNew("tcga-brca", matrix_data, rowOrder = rowOrder, colOrder = colOrder)

Hierarchical Clustering Options

For hierarchical clustering, additional options are available to specify the distance measure and clustering method.

The arguments to chmNew() ‘rowDist’ and ‘colDist’ specify the distance measure for rows and columns, respectively. The available choices include those of the stats::dist() function, as well as two additional distance metric options: ‘cosine’ and ‘correlation’. The default choice is ‘correlation’, which computes the distance measure as 1 minus the Pearson correlation among the rows/columns.

Arguments ‘rowAgglom’ and ‘colAgglom’ specify the clustering method, and the available options are those of the stats::hclust function. The default is ‘ward.D2’.

This example specifies an NG-CHM with hierarchical clustering using the Euclidean distance metric and complete linkage clustering algorithm for both rows and columns:

hm <- chmNew("tcga-brca", matrix_data, rowDist = "euclidean", colDist = "euclidean", rowAgglom = "complete", colAgglom = "complete")

Back to top

Explicit Clustering

Clustering may be performed explicitly (e.g. via stats::hclust()) and used in constructing an NG-CHM. The following creates an object of class hclust from the distance matrix of the demo expression data, for both the rows and columns, and uses those clustering results to construct an NG-CHM.

rowClust <- stats::hclust(dist(matrix_data))
colClust <- stats::hclust(dist(t(matrix_data)))
hm <- chmNew("tcga-brca", matrix_data, rowOrder = rowClust, colOrder = colClust)

Similarly, an hclust object can be transformed into a dendrogram and used for the row and/or column ordering. The example below uses the clustering results from above as dendrograms:

rowDendrogram <- as.dendrogram(stats::hclust(dist(matrix_data)))
colDendrogram <- as.dendrogram(stats::hclust(dist(t(matrix_data))))
hm <- chmNew("tcga-brca", matrix_data, rowOrder = rowDendrogram, colOrder = colDendrogram)

Back to top