R/plot_nnls_solution.R

In eyzhao/SignIT: What the Package Does (One Line, Title Case)

#' Plots a simple NNLS solution
#'
#' Obtains the naive NNLS signature solution and plots the exposures.
#'
#' Non-negative least squares computes the value of x which minimizes the residual sum
#' of squares \eqn{\lVert Ax = b \rVert^2} given a known matrix A and vector b.
#' This provides a simple solution to the mutation signature problem. However,
#' merely minimizing the residual sum of squares may result in overfitting the solution.
#' This plotting function computes and plots a simple NNLS-based solution to the
#' exposure vector and returns a plot. This can serve as a baseline comparison or sanity check
#' against the MCMC methods featured in SignIT.
#'
#' @param exposures_mcmc_output
#'
#' @return A ggplot object showing the NNLS solution for each exposure value.
#'
#' @import nnls
#' @import tibble
#' @import dplyr
#' 
#' @export

plot_nnls_solution <- function (exposures_mcmc_output, signature_trim = "Signature") 
{
  nnls_solution <- nnls(
    exposures_mcmc_output$reference_signatures %>%
      reference_signatures_as_matrix(exposures_mcmc_output$mutation_catalog),
    exposures_mcmc_output$mutation_catalog$count
  )$x
  nnls_plot <- tibble(signature = exposures_mcmc_output$signature_names %>% 
                        factor(levels = exposures_mcmc_output$signature_names), 
                      nnls = nnls_solution) %>% mutate(signature = trim_signature_names(signature, 
                                                                                        signature_trim)) %>% ggplot(aes(x = signature, y = nnls)) + 
    geom_point() + rotate_x_axis_labels()
}