Visualize Clusters Using t-SNE or UMAP

Generates 2D visualization of high-dimensional data using either t-SNE or UMAP dimensionality reduction, with options for cluster labeling and aesthetic customization.

Usage

DrawCluster(
  data,
  label = NULL,
  point_size = 1,
  method = c("tsne", "umap"),
  draw_cluster_text = TRUE,
  calculated = TRUE,
  pca = TRUE,
  perplexity = 100,
  plot = TRUE,
  seed = 1
)

Arguments

data: Input data matrix (features x samples) or pre-computed coordinates if calculated=FALSE
label: Vector of cluster labels for each sample (optional)
point_size: Size of points in the plot (default: 1)
method: Dimensionality reduction method: "tsne" or "umap" (default: "tsne")
draw_cluster_text: Whether to display cluster labels at median positions (default: TRUE)
calculated: Whether to compute dimensionality reduction (TRUE) or use provided coordinates (FALSE)) (default: TRUE)
pca: Whether to perform PCA preprocessing for t-SNE (default: TRUE)
perplexity: t-SNE perplexity parameter (default: 100)
plot: Whether to display the plot (default: TRUE)
seed: Random seed for reproducibility (default: 1)

Value

A list containing:

p: ggplot2 object of the visualization
x: Data frame with coordinates (V1 and V2 columns)
cell_group: Factor vector of cluster labels

Details

Key features:

Supports both t-SNE (t-Distributed Stochastic Neighbor Embedding) and UMAP (Uniform Manifold Approximation and Projection)
Option to use pre-computed coordinates or calculate new embeddings
Automatic cluster labeling at median positions
Customizable plotting parameters and reproducible results via seed setting

The function returns both the plot object and the coordinates for further analysis.

Author

Bin Duan (binduan\@sjtu.edu.cn)

Examples

if (FALSE) { # \dontrun{
# Using example data from scLearn package
data(QueryCellData)

# Get normalized expression matrix
norm_expr <- logcounts(QueryCellData)
cell_types <- colData(QueryCellData)$cell_type1

# t-SNE visualization
tsne_res <- DrawCluster(
  data = norm_expr,
  label = cell_types,
  method = "tsne",
  perplexity = 30,
  point_size = 1.5
)

# UMAP visualization without cluster labels
umap_res <- DrawCluster(
  data = norm_expr,
  label = cell_types,
  method = "umap",
  draw_cluster_text = FALSE,
  point_size = 2
)

# Using pre-computed coordinates
precomputed_coords <- data.frame(V1 = rnorm(ncol(norm_expr)),
                               V2 = rnorm(ncol(norm_expr)))
DrawCluster(
  data = precomputed_coords,
  label = cell_types,
  calculated = FALSE
)
} # }