Skip to contents

Compute PCA embeddings from a raw counts matrix

Source: R/utils_pca.R
do_pca.Rd

First, log-normalizes and Z-scores the counts matrix and then performs PCA using SVD.

Usage

do_pca(counts, npcs)

Arguments

counts

A n_genes x n_cells counts matrix. Must be convertible to dgCMatrix.

npcs

Number of PCs to compute.

Value

A list with two features:

  • loadings: A n_genes x npcs matrix of gene loadings for each PC. Each column is a unit vector.

  • embeddings: A n_cells x npcs matrix of cell embeddings across all PCs. Each column j has magnitude equal to the jth singular value. That is, PCs with larger contribution to the total variance will have embeddings of proportionally larger magnitude.