R/post_processing.R

In sipss/AlpsLCMS: Signal processing of LC-MS metabolomics

#' Metabolite Candidates
#'
#' Post processing after obtaining the list of potential metabolite candidates.
#' These metabolites can be filtered by biofluid and are sorted by p-value in
#' ascending order. Unknown metabolites can be removed. After that, the results
#' the results are plotted using a word cloud. The lower the p-value, the higher
#' the font size of the metabolite. Additionally, we plot the a bar plot
#' corresponding to the number of metabolites candidates associated to a single
#' compound.
#'
#' @param path_result The folder where the .csv file where the annotated table
#'   of metabolites is stored
#' @param biofluid_type Filters the results by biofludid type.
#' @param significance Filters the results by significance.
#' @param metabolite_rm Logical. Remove "Unknown" metabolites.
#' @param  max_words  Maximum number of words allowed in the word cloud.
#' @return Two plots: A word cloud with the potential identified metabolites and
#'   a bar plot with the number of potential cantidates by detected compound.
#' @family data visualization functions
#' @export
#' @examples
#' \dontrun{
#' path_result <-  system.file("extdata",
#'                             "results_project",
#'                             "tables",
#'                             "metaboliteTable.csv",
#'                             package = "AlpsLCMS")
#' plots <- lcms_plot_metabolites(path_result, biofluid_type = "Any",
#'                                    significance = 0.05,
#'                                    metabolite_rm = FALSE,
#'                                    max_words = 250)
#' plots$cloud
#'
#' plots$freq
#' }

lcms_plot_metabolites <- function(path_result, biofluid_type = "Any",
                                  significance = 0.05,
                                  metabolite_rm = TRUE,
                                  max_words = 250){
  spectra <- NULL
  rt <- NULL
  Name <- NULL
  p.adj <- NULL
  Metabolite <- NULL
  Biofluid <- NULL
  Retention_Time <- NULL
  Pvalue_Adj <- NULL
  Angle <- NULL
  Spectra <- NULL
  Selected_Biofluid <- NULL
  n <- NULL

  #Load Measured Data
  suppressWarnings(Measured <- readr::read_csv(file = path_result,
                                               col_names = TRUE,
                                               col_types = readr::cols_only(spectra = readr::col_number(),
                                                                            rt = readr::col_double(),
                                                                            Name = readr::col_factor(),
                                                                            Biofluid = readr::col_character(),
                                                                            p.adj = readr::col_double()),
                                               skip_empty_rows = TRUE))
  attr(Measured, "spec") <- NULL

  # Some Checks

  if (!is.logical(metabolite_rm)){
    stop("The variable Metabolite_rm must be of class logical")
  }


  Biofluid_names <- Measured %>%
    purrr::pluck("Biofluid") %>%
    stringr::str_split( ";")  %>%
    base::unlist() %>%
    stringr::str_replace("^ ", "") %>%
    base::unique() %>%
    base::sort()


  if (!is.character(biofluid_type)){
    stop("The variable Biofluid_type must be of class character")
  } else if (!(biofluid_type %in% Biofluid_names)){
    if (biofluid_type != "Any"){
      stop(paste0("Your selected Biofluid is not among the following:",
                  "\n",stringr::str_c(Biofluid_names, collapse = ", "), "."))
    }

  }


  if (!is.null(significance)){
    if(is.numeric(significance)){
      if((significance > 1) | (significance < 0)){
        stop("Significance must be set to a value in the range [0, 1]")
      }
    }else {
      stop("Significance must be a variable with class either numeric or NULL")
    }
  }

  # Modify Column_names and drop na

  Measured <- Measured %>%
    dplyr::rename(Spectra = spectra,
                  Retention_Time = rt,
                  Metabolite = Name,
                  Pvalue_Adj = p.adj) %>%
    tidyr::drop_na()


  # Remove unknown Metabolites if needed

  if(metabolite_rm == TRUE){
    Measured <- Measured %>%
      dplyr::filter(Metabolite != "Unknown")
  }


  # Filter by Biofluid

  if(biofluid_type == "Any"){
    Measured <- Measured %>%
      dplyr::mutate(Selected_Biofluid = "Any")

    max_area_wordcloud <- 11

  } else{
    Measured <- Measured %>%
      dplyr::filter_at(dplyr::vars(Biofluid), dplyr::all_vars(stringr::str_detect(Biofluid, biofluid_type))) %>%
      dplyr::mutate(Selected_Biofluid = biofluid_type)

    max_area_wordcloud <- 11
  }

  # Select and arrange the data

  Measured <- Measured %>%
    dplyr::select(Retention_Time, Spectra,  Metabolite, Selected_Biofluid, Biofluid, dplyr::everything()) %>%
    dplyr::arrange(Retention_Time, Spectra)

  #Filter by p-value

  if(!is.null(significance)){
    Measured <- Measured %>%
      dplyr::filter(Pvalue_Adj <= significance)
  }


  # plots
  #word_cloud
  plot_df <- Measured %>%
    dplyr::select(Retention_Time, Spectra, Metabolite, Pvalue_Adj) %>%
    dplyr::mutate(Spectra = as.factor(Spectra)) %>%
    dplyr::mutate(Angle = 0 * sample(c(0, 1),#90
                                     dplyr::n(), replace = TRUE,
                                     prob = c(60, 40))) %>%
    dplyr::top_n(max_words,-log10(Pvalue_Adj))


  #
  text_title_cloud <- paste0("Metabolite Word Cloud. Biofluid: ",
                             base::unique(Measured$Selected_Biofluid), ". ","Pvalue < 0.05.")

  set.seed(42)
  cloud <- ggplot2::ggplot(
    plot_df,
    ggplot2::aes(
      label = Metabolite, size =  -log10(Pvalue_Adj),
      colour = Retention_Time,
      angle = Angle
    )
  ) +
    ggwordcloud::geom_text_wordcloud_area() +
    ggplot2::scale_size_area(max_size = max_area_wordcloud) +#8 guay export, 4 plot
    ggplot2::theme_minimal() +
    ggplot2::scale_colour_continuous(type = "viridis") +
    ggplot2::ggtitle(text_title_cloud)

  # plots
  #Barplot:  Spectra - Metabolite Candidate

  freq <- plot_df %>%
    dplyr::group_by(Spectra) %>%
    dplyr::count() %>%
    ggplot2::ggplot(ggplot2::aes(x = stats::reorder(Spectra, n),
                                 y = n, fill = stats::reorder(Spectra, n),
                                 color = stats::reorder(Spectra, n))) +
    ggplot2::geom_bar(stat="identity", position = "stack", alpha = 0.7, size = 1) +
    ggplot2::scale_x_discrete("Spectrum Number") +
    ggplot2::scale_y_continuous("Number of Metabolite Candidates") +
    ggplot2::theme_minimal() +
    ggplot2::theme(axis.ticks = ggplot2::element_blank(),
                   plot.title = ggplot2::element_text(size = 17),
                   axis.text.x = ggplot2::element_text(angle = 90),
                   axis.text = ggplot2::element_text(size = 8, color = "black"),
                   axis.title = ggplot2::element_text(size = 14, color = "black"),
                   axis.line = ggplot2::element_line(color = "black",
                                                     size = 1, linetype = "solid"),
                   legend.position = "none") +
    ggplot2::ggtitle("Metabolite Candidates for Spectrum")

  plot_list <- list(cloud = cloud, freq = freq)
  plot_list

}