Skip to contents

Baseline Dimension Reduction of RNA-seq Data

baseline_dimreduction.Rd

Takes a formatted DESeqDataSet object and performs several exploratory analyses to evaluate sample similarity and major sources of variation within the dataset. Performs baseline exploratory dimension reduction and sample-level visualization of a DESeqDataSet. Variance-stabilized expression values are used to calculate sample distances and principal components. The most variable genes are used for PCA and downstream visualization.

The analysis:

  1. applies a variance-stabilizing transformation to the expression matrix,

  2. identifies the top variable genes,

  3. calculates Euclidean distances between samples,

  4. performs principal component analysis (PCA) using the top variable genes,

  5. generates sample embedding plots,

  6. generates an elbow plot showing the variance explained by the leading principal components, and

  7. performs UMAP for nonlinear dimensionality reduction.

Usage

baseline_dimreduction(bulk_dataset, top_var = 2000, grouping = NA)

Arguments

bulk_dataset

A formatted DESeqDataSet (DDS) object containing the RNA-seq expression data and associated sample metadata.

top_var

Integer specifying the number of genes with the highest variance to retain for PCA and downstream dimension reduction. For example, top_var = 2000 uses the 2,000 most variable genes.

grouping

Character string specifying the sample metadata variable used to annotate or group samples in the distance and dimension reduction visualizations. The variable should correspond to a column in the colData of bulk_dataset.

Value

  • distance_plot: euclidean distance plot,

  • elbow_plot: shows which PC has the majority of the variance within the data

  • PCA plot: scatterplot of the left unitary vectors

  • UMAP_plot: A UMAP visualization of the sample embeddings.

    results$distance_plot results$PCA_plot results$elbow_plot results$umap_plot

Examples

if (FALSE)  baseline_dimreduction(dds_object, top_var = 2000, grouping = NA) # \dontrun{}