PCA prediction by projecting query/test samples onto PCA of reference/train samples

Plot titles are automatically constructed based on parameters. Adding a savedir argument will result in intermediate plots and final projected scores to be output.

Usage

predict_PCA(
  train,
  test,
  scale = FALSE,
  varimax = FALSE,
  train_name = "Train",
  train_anno = NULL,
  train_annoname = NULL,
  train_annotype = NULL,
  train_colors = NULL,
  test_name = "Test",
  test_anno = NULL,
  test_annoname = NULL,
  test_annotype = NULL,
  test_colors = NULL,
  ellipse = FALSE,
  label = FALSE,
  flip = FALSE,
  savedir = NULL,
  height = 8,
  width = 8,
  fmt = "png",
  test_only_plt = FALSE,
  rank = 3
)

Arguments

train

string/df; filepath to/df of reference/train samples data, genes/features as rownames, samples/observations as colnames

test

string/df; filepath to/df of query/test samples data, genes/features as rownames, samples/observations as colnames

scale

logical; T to scale variables to unit variance

varimax

logical; T to perform varimax rotation on train + test

train_name

string; descriptor for train samples

train_anno

df; annotation info for train samples

train_annoname

string; colname in train_anno matching point name

train_annotype

string; colname in train_anno with info to color by

train_colors

char vector; list of colors, length = # of unique train_annotype, set NULL for gray

test_name

string; descriptor for test samples

test_anno

df; annotation info for test samples

test_annoname

string; colname in test_anno matching point name

test_annotype

string; colname in test_anno with info to color by

test_colors

char vector; list of colors, length = # of unique test_annotype

ellipse

logical (vector); if length 2, ellipse[1] for train data ellipse, ellipse[2] for test data ellipse

label

logical (vector); if length 2, label[1] for train data label, label[2] for test data label

flip

NOT IMPLEMENTED - logical (vector); if length 2, flip[1] to flip x-axis values, flip[2] to flip y-axis values

savedir

string; directory (+ prefix) to save output under; if directory, end string with "/"

height

numeric; plot height

width

numeric; plot width

fmt

string; plot output format (ex. "png", "pdf")

test_only_plt

logical; save scatter of projected test samples only

rank

integer; maximal # of PCs to be used in prcomp; for compatibility w/ glab version

Value

ggplot object

See also

run_PCA(), plot_PCA(), rotate_varimax()

Examples

library(dplyr)
data(iris)
iris$Sample <- rownames(iris)
set.seed(13)
samp <- sample(nrow(iris), nrow(iris)*0.75) # Train data is 75% of iris
iris_train <- iris[samp,] %>%
  mutate(Batch = "Train",
         Species = paste0(Species, "_Train"))
iris_test <- iris[-samp,] %>%
  mutate(Batch = "Test",
         Species = paste0(Species, "_Test"))
Rubrary::predict_PCA(
  train = t(iris_train[,1:4]),
  test = t(iris_test[,1:4]),
  train_name = "Iris Train", train_anno = iris_train[,c("Sample", "Species")],
  train_annoname = "Sample", train_annotype = "Species",
  train_colors = c("pink", "red", "darkred"),
  test_name = "Iris Test", test_anno = iris_test[,c("Sample", "Species")],
  test_annoname = "Sample", test_annotype = "Species",
  test_colors = c("lightskyblue", "blue", "navy"),
  ellipse = c(FALSE, TRUE), label = FALSE
)
#> Common features: 4