R/PCAvisualization.R

In bcaffo/MRIcloudT1volumetrics: MRIcloud Analysis of T1 Volumetric Output

#' @name PCAvisualization
#' @title A function for visualization of principal component rotations
#' @author Luchao Qi
#' @param pca.result should use function MRIPCA to get a list of pca result
#' @param PC index of PC, default = 1
#' @import ggplot2
#' @importFrom ggplot2 aes
#' @importFrom dplyr mutate
#' @importFrom tidyr gather
#' 
#' @export

PCAvisualization = function(pca.result, PC = 1){
  Age = roi = Comp = id = value = NULL
  rm(list = c("Age", "roi", "Comp", "id", "value"))
  
  pca.rotations = sapply(pca.result, function(i) i$rotation)
  dat_visualization = lapply(pca.rotations, function(i){
    df = prop.table(abs(i),margin = 2) %>%
      as.data.frame()
    df = df %>% 
      mutate(id = 1:nrow(df))
    df = df %>% 
      tidyr::gather(key = roi, value = Comp, -id)
    df %>% 
      filter(Comp == paste0('PC',PC)) %>% 
      select(c(roi, value))
  }) %>%
    do.call(what = "rbind") #convert to data frame
  dat_visualization = dat_visualization %>%
    mutate(Age = sapply(strsplit(rownames(dat_visualization), "[.]"),function(x) x[1])
    ) # rownames reformat
  
  
  ggplot(dat_visualization, aes(x = Age,y = value,colour = roi,group = roi)) +
    geom_line() +
    labs(x = 'Age', y = 'Weight', title = paste0('Compositional analysis: weight of ROIs in PC',PC))
}