get_pcs: Get PCs

In marcalva/diem: Debris-Containing Droplet Identification using EM

Get PCs

Description

Run PCA and get top n_pcs PCs for the test set. The PCs are used as the features for the initial k-means clustering. Only droplets with at least min_genes are used in the PCA, and thus used in the initialization. The counts data for the test set are count-normalized to the median and log transformed. Then the top n_var_genes variable genes are calculated using the function get_var_genes. PCA is run on the normalized count data for these variable genes only.

Usage

get_pcs(x, droplets.use = NULL, min_genes = 200, n_var_genes = 2000,
  lss = 0.3, threads = 1, n_pcs = 30, seedn = 1)

Arguments

x	An SCE object.
droplets.use	Specify droplets to calculate PCs for.
min_genes	Calculate PCs from droplets with at least this many genes detected.
n_var_genes	Number of top variable genes to use for PCA.
lss	The span parameter of the loess regression, the parameter for the function loess. The loess regression is used to regress out the effect of mean expression on variance.
threads	Number of threads for parallel execution. Default is 1.
n_pcs	Number of PCs to return.
seedn	The seed to set for irlba PCA calculation. It is set to 1 for reproducibility but can be set to NULL for a random initialization.

Value

An SCE object with PCs

Examples

# Get PCs with default parameters
sce <- get_pcs(sce)

# Run initialization with droplets that have at least 150 genes
# detected
sce <- get_pcs(sce, min_genes = 150)

# Using top 3,000 variable genes
sce <- get_pcs(sce, n_var_genes = 3000)

# Use top 50 PCs for initialization
sce <- get_pcs(sce, n_pcs = 50)

# Return PCs from random irlba initializations
sce <- get_pcs(sce, seedn = NULL)
sce <- get_pcs(sce, seedn = NULL)
sce <- get_pcs(sce, seedn = NULL)

marcalva/diem documentation built on Jan. 1, 2023, 2:33 a.m.