identify significant prinicipal components (PCs)
signPCA(
gobject,
name = "pca",
method = c("screeplot", "jackstraw"),
expression_values = c("normalized", "scaled", "custom"),
reduction = c("cells", "genes"),
pca_method = c("irlba", "factominer"),
rev = FALSE,
genes_to_use = NULL,
center = T,
scale_unit = T,
ncp = 50,
scree_ylim = c(0, 10),
jack_iter = 10,
jack_threshold = 0.01,
jack_ylim = c(0, 1),
verbose = TRUE,
show_plot = NA,
return_plot = NA,
save_plot = NA,
save_param = list(),
default_save_name = "signPCA"
)
giotto object
name of PCA object if available
method to use to identify significant PCs
expression values to use
cells or genes
which implementation to use
do a reverse PCA
subset of genes to use for PCA
center data before PCA
scale features before PCA
number of principal components to calculate
y-axis limits on scree plot
number of interations for jackstraw
p-value threshold to call a PC significant
y-axis limits on jackstraw plot
verbosity
show plot
return ggplot object
directly save the plot [boolean]
list of saving parameters from all_plots_save_function()
default save name for saving, don't change, change save_name in save_param
ggplot object for scree method and maxtrix of p-values for jackstraw
Two different methods can be used to assess the number of relevant or significant
prinicipal components (PC's).
1. Screeplot works by plotting the explained variance of each
individual PC in a barplot allowing you to identify which PC provides a significant
contribution (a.k.a. 'elbow method').
2. The Jackstraw method uses the permutationPA
function. By
systematically permuting genes it identifies robust, and thus significant, PCs.