Skip to contents

cluster_metrics: Calculates clustering metric scores

cluster_metrics.Rd

Uses the ClusterSim package to calculate metrics for cluster consistency. These metrics do not require a priori knowledge of the truth set, which makes them ideal for scRNA clustering.

Usage

cluster_metrics(
  so,
  cluster_list,
  dims = 1:20,
  reduction = "pca",
  silhouette = TRUE
)

Arguments

so

A Seurat single cell RNA (scRNA) object

cluster_list

A vector of cluster resolution names (e.g. "SCT_snn_res.0.2")

dims

An integer vector of the principal components to be used in calculating the cluster metrics. May not exceed the number of principal components available

reduction

A string for the reduction to be used in calculating scores

silhouette

A Boolean for whether to include silhouette scoring

Value

Returns a table of clustering metrics for each cluster resolution selected

Details

Calculates three consistency scores to evaluate clustering effectiveness: Calinski-Harabasz: Lower scores preferred. Davies-Bouldin: Higher scores preferred. Silhouette score: Higher scores preferred. Recommended to only run when data has fewer than 90,000 data points