In steverozen/mSigHdp: Mutational signature discovery using HDP (hierarchical Dirichlet process)

require(ggplot2, quietly = TRUE)

mSigHdp can downsample input spectra to reduce mutation counts in more highly mutated tumors, while leaving the mutational spectra with lower mutation counts as they are. The operational definition of "highly" or "lower" is governed by argument \code{thres}.

Downsampling reduces running time for burnin and Gibbs sampling.

x_vec <- seq(0, 2e+04, 100)
thres_vals <- c(1000, 3000, 5000, 10000)

y_df <- 
  do.call(rbind, 
          lapply(
            X = thres_vals,
            function(thres_val) {
              data.frame(thres = thres_val, 
                         x     = x_vec, 
                         y     = mSigHdp::downsample(x_vec, thres = thres_val))
              }))

ggObj <- 
  ggplot2::ggplot(y_df) + xlab("Counts") + ylab("Downsampled counts") +
  ylim(0, 12000) +
  geom_point(mapping = aes(x = x, y = y, color = as.factor(thres)), 
             size = 0.05) +
  guides(color = guide_legend(title = "thres"))

for(thres_val in thres_vals){
  ggObj <- ggObj +
    geom_hline(yintercept = thres_val, size = 0.25) 
}

ggObj

version

steverozen/mSigHdp documentation built on Feb. 6, 2023, 1:36 a.m.