compute_pcas: Compute principal components.
In mirvie/mirmodels: Models Built by the Informatics Team at Mirvie

Description Usage Arguments Value Examples

Computation can be done using the recipes package (robust = FALSE, uses regular PCA) or with the rrcov package which uses rrcov::PcaGrid(), a robust PCA computation with built-in outlier identification.

compute_pcas(
  df,
  num_comp = 5,
  subset = NULL,
  normalize = TRUE,
  robust = FALSE,
  crit_pca_distances = 0.999
)

`df`	A data frame.
`num_comp`	A positive integer. The number of PCA components to compute.
`subset`	A character vector. The subset of columns to use for PCA calculation. Default is all columns. Columns in `df` outside of this subset will be kept in the result.
`normalize`	A flag. Center and scale before PCA calculation?
`robust`	A flag. Use robust PCA (from `rrcov::PcaGrid()`)?
`crit_pca_distances`	A number between 0 and 1. Outlier identification parameter. See `rrcov::PcaGrid()`. The larger this number, the fewer outliers identified.

A data frame with the principal components with attributes

var_exp: A numeric vector. The variance explained by each component.
loadings: A data frame. The contribution of each variable to each principal component.
outlier: A logical vector with length nrow(df). TRUE for outlier, FALSE otherwise. This is present with robust = TRUE only.

if (rlang::is_installed("mirmisc")) {
  ga_data <- get_ga_data(
    log2 = TRUE,
    gene_predicate = ~ median(.) > 0
  )
  pca_reg <- compute_pcas(ga_data,
    subset = intersect(
      mirmisc::get_gene_names(),
      names(ga_data)
    ),
    robust = FALSE
  )
  pca_rob <- compute_pcas(ga_data,
    subset = intersect(
      mirmisc::get_gene_names(),
      names(ga_data)
    ),
    robust = TRUE
  )
  ggplot2::ggplot(
    pca_reg,
    ggplot2::aes(PC1, PC2, color = meta_collectionga)
  ) +
    ggplot2::geom_point() +
    ggplot2::scale_color_viridis_c() +
    ggplot2::ggtitle("GAPPS regular")
  ggplot2::ggplot(
    pca_rob,
    ggplot2::aes(PC1, PC2, color = meta_collectionga)
  ) +
    ggplot2::geom_point() +
    ggplot2::scale_color_viridis_c() +
    ggplot2::ggtitle("GAPPS robust")
}