metamoRph

Takes in a count matrix (where genes (features) are rows and samples are columns) as well as a named vector with the eigenvalues (see run_pca()) and pulls the gene (feature) information from the rotation vector and cuts down the new_counts matrix to match the rotation vector gene (feature) names. Any genes (features) missing from the input new_counts matrix will be replaced with zeros.

Usage

metamoRph(
  new_counts,
  rotation,
  center_scale = NULL,
  normalization = TRUE,
  sample_scale = "cpm",
  feature_scale = FALSE,
  log1p = TRUE
)

Arguments

new_counts

raw gene count matrix (where genes are rows and samples are columns)

rotation

matrix where the row names are genes and the col names are the principal components. If you used metamoRph::run_pca() then this would be in the output$PCA$rotation slot.

center_scale

list object where the $center slot has the center values and the $scale slot has the scale value for the "scale" function. If you do not give a value here, then feature/gene scaling WILL NOT HAPPEN.

normalization

Default is TRUE, if set to FALSE will override sample_scale and log1p and not do any sample scaling

sample_scale

Options include cpm, seurat, zscale, and none. This MUST match the normalization used in the data for run_pca (or where your rotation came from).

feature_scale

Default is FALSE, if TRUE will apply feature (gene) scaling to the input (query) data (if center_scale is left empty). VERY DANGEROUS as this will likely put the query data in a different scale than your reference data. Use with deliberate intent.

log1p

Default is TRUE. Again, must match the normalization used in the data for run_pca (or where your rotation came from).

Value

A matrix with the transformated eigenvalue matrix which should be equivalent to the original rotation matrix's eigenvalue/pattern matrix (The $x slot from the output of prcomp)

Details

The function will scale the new_counts matrix in the same manner as run_pca() and matrix multiply by the rotation vector. The output is equivalent to the prcomp "$x" matrix.