comparison Testtest/GCMS-test_output.R @ 0:40de28c7d3fb draft

Uploaded

0:40de28c7d3fb

# author: Pauline Ribeyre


#####################
# required packages #
#####################

library("parallel") # provides cluster methods and "parLapplyLB" function
library("grDevices")  # provides jpeg handling methods


############
#   data   #
############

update_list_of_file_names <- function() {
  # Writes in a file the list of file names using the files created previously in ./RDatas.
  
  file.create(source_file_names, showWarnings = FALSE) # erase the file
  directory <- "RDatas"
  files <- list.files(directory, full.names = TRUE)
  
  for (f in files) {
    f <- substr(f, 8, nchar(f) - 6)
    write(f, file = source_file_names, append = TRUE)
  }
  
}


titles_to_columns <- function(indicators) {
  # Parse the "title" column of the indicator dataframe to separate the different parameters and their values.
  #
  # Args:
  #   indicators: dataframe (one row per test) containing the results.
  #
  # Returns:
  #   new_indicators: copy of "indicators" with one column added for each parameter that varied.
  
  default_settings <- FALSE
  
  # parse the title
  param_names <- c()
  for (title in indicators$title) {
    
    if (title == "default_settings") {
      default_settings <- TRUE
      break
    }
    
    else {
      params <- strsplit(title, "_")[[1]]
      for (param in params) {
        name <- strsplit(param, "=")[[1]][1]
        
        # some lowercase values create errors because they are primitive functions
        substr(name, 1, 1) <- toupper(substr(name, 1, 1))
        
        if (!name %in% param_names)
          param_names <- c(param_names, name)
        value <- strsplit(param, "=")[[1]][2]
        if (!exists(name))
          assign(name, c())
        assign(name, c(get(name), value))
      }
    }
    
  }
  
  new_indicators <- indicators
  
  # add the columns to the dataframe
  if (!default_settings) {
    for (name in param_names) {
      new_indicators <- cbind(new_indicators, get(name))
      names(new_indicators)[ncol(new_indicators)] <- name
    }
    
    # order the dataframe's columns
    order <- c(1, (ncol(new_indicators) - length(param_names) + 1) : ncol(new_indicators), 2 : (ncol(new_indicators) - length(param_names)))
    new_indicators <- new_indicators[, order]
  }
  
  return (new_indicators)
  
}


create_summary <- function(nb_cores, count_duplicates = FALSE) {
  # Reads the files created by runGC_vary_parameters() and calculates the quality criteria for each test.
  #
  # Args:
  #   nb_cores: maximum number of cores to use.
  #   count_duplicates: calculate the number of count_duplicates obtained by each test (slow).
  #
  # Returns:
  #   A dataframe (one row per test) containing the results:
  #     title, nb ions, nb zeros/line, nb ions/intensity range, presence of valine, nb count_duplicates (opt).
# 
#   # Returns:
#   #   A list containing:
#   #   1: dataframe (one row per test) containing the results:
#   #     title, nb ions, nb zeros/line, nb ions/intensity range, presence of valine, nb count_duplicates (opt).
#   #   2: list of the settings sets for each test.
  
  time.start <- Sys.time() # start the timer
  
  file_names <- readLines(source_file_names)
  
  dir.create("count_duplicates", showWarnings = FALSE) #create the folder where the details of the count_duplicates will be saved
  
  # summ <- create_summary_parallel(1, file_names)
  # cat("indic:\n",summary[1][[1]],"\n")
  
  # run the function on several cores
  if (length(file_names) != 0) {
    if (length(file_names) < nb_cores)
      nb_cores <- length(file_names)
    cluster <- makeCluster(nb_cores) #, outfile = "")
    summaries <- parLapplyLB(cluster, 1:length(file_names), create_summary_parallel, file_names = file_names, count_duplicates = count_duplicates) #, pb = pb)
    stopCluster(cluster)
  }
  else
    stop("There are no files to generate the output from.")
  
  # concatenate the results obtained by all the cores
  indicators <- NULL
  for (summary in summaries)
    indicators <- rbind(indicators, t(summary))
  # settings_list <- c()
  # for (summary in summaries) {
  #   settings <- summary[2][[1]]
  #   indic <- summary[1][[1]]
  # 
  #   indicators <- rbind(indicators, t(indic))
  #   settings_list <- c(settings_list, settings)
  # }
  indicators <- data.frame(indicators)

  col_names <- c("title", "nb_pseudospectra", "zeros_per_line", "f0to10E3", "f10E3to10E4", "f10E4to10E5",
                 "f10E5to10E6", "f10E6to10E7", "f10E7")
  
  if (check_ions)
    col_names <- c(col_names, ions_name)
  
  names(indicators)[1:length(col_names)] <- col_names
  if (count_duplicates)
    names(indicators)[(ncol(indicators) - 1):ncol(indicators)] <- c("count_duplicates", "nb_ions")
  
  indicators <- titles_to_columns(indicators)

  time.end <- Sys.time() # stop the timer
  Tdiff <- difftime(time.end, time.start)
  print(Tdiff)
  
  # return (list(indicators, settings_list))
  return (indicators)
  
}


create_summary_parallel <- function(n, file_names, count_duplicates = FALSE) { #}, pb) {
  # Reads the files created by runGC_vary_parameters() and calculates the quality criteria for each test.
  #
  # Args:
  #   n: index of the current test, to select the corresponding title.
  #   file_names:  list of titles (one for each test) (concatenation of the values taken by the varied parameters).
  #   nb_cores: maximum number of cores to use.
  #   count_duplicates: calculate the number of count_duplicates obtained by each test (slow).
  #
  # Returns:
  #   A dataframe (one row per test) containing the results:
  #     title, nb ions, nb zeros/line, nb ions/intensity range, presence of valine.
  # 
  # # Returns:
  # #   A list containing:
  # #   1: dataframe (one row per test) containing the results:
  # #     title, nb ions, nb zeros/line, nb ions/intensity range, presence of valine.
  # #   2: list of the settings sets for each test.
  
  source(source_spectrum, environment())
  
  intensities_x <- c(1000, 10000, 100000, 1000000, 10000000)
  
  this_title <- file_names[n]
  
  # calculate the number of zeros
  file_title <- paste0("Peak_tables/", this_title, ".tsv")
  peak_table <- read.table(file_title, sep="\t", header=TRUE)
  peak_table_values <- peak_table[,5:ncol(peak_table)]
  zeros <- sum(peak_table$nb_zeros)
  nb_lines <- nrow(peak_table)
  zeros_per_line <- round(zeros/nb_lines, 4)
  
  # count the number of ions by intensity range
  intensities_y <- c()
  intensities_y[1] <- length(rowMeans(
    peak_table_values, na.rm = TRUE)[rowMeans(peak_table_values, na.rm = TRUE) < intensities_x[1]])
  for (i in 2:length(intensities_x))
    intensities_y[i] <- length(rowMeans(
      peak_table_values, na.rm = TRUE)[rowMeans(peak_table_values, na.rm = TRUE) < intensities_x[i] &
                                       rowMeans(peak_table_values, na.rm = TRUE) > intensities_x[i - 1]])
  intensities_y[i + 1] <- length(rowMeans(
    peak_table_values, na.rm = TRUE)[rowMeans(peak_table_values, na.rm = TRUE) > intensities_x[i]])
  
  # load the settings
  file_title <- paste0("RDatas/", this_title, ".RData")
  load(file_title)

  # count the number of count_duplicates and record in a file
  if (count_duplicates) {
    file_title <- paste0("count_duplicates/", this_title, ".tsv")
    nb_count_duplicates <- data.frame(count_duplicates_function(GC_results))
    names(nb_count_duplicates) <- c("mz_min", "mz_max", "rt", "count_duplicates")
    write.table(nb_count_duplicates, file = file_title, sep = "\t", row.names = FALSE)
    count_duplicates <- nrow(nb_count_duplicates)
    
    nb_ions <- 0
    pseudospectra <- GC_results$PseudoSpectra
    for (ps in pseudospectra) {
      nb_ions <- nb_ions + nrow(ps$pspectrum)
    }
  }
  
  summary <- c(this_title, nb_lines, zeros_per_line, intensities_y)
  
  # check the presence of ions
  if (nb_ions_to_check > 0) {
    for (i in 1:nb_ions_to_check) {
      name <- ions_name[[i]]
      rt <- ions_rt[[i]]
      mzs <- ions_mzs[[i]]
      cat("Check:", name, rt, mzs, "\n")
      
      value <- is_ion_present(GC_results, rt = rt, mz_list = mzs)
      assign(name, value)
      summary <- c(summary, get(name))
    }
  }
  
  # check the presence of ion valine 12C and 13C
  # valine <- is_ion_present(GC_results, rt = 9.67, mz_list = list(218.1105, 219.1146))

  # summary <- c(summary, valine)
  
  if (count_duplicates)
    summary <- c(summary, count_duplicates, nb_ions)

  # return (list(summary, settings))
  return (summary)
  
}


############
#  graphs  #
############

ions_per_intensity <- function(indicators) {
  # For each test, plots the number of ions for each range of intensity.
  #
  # Args:
  #   indicators: dataframe (one row per test) containing the results. 
  
  pdf(intensity_graph_out)
  par(mar = c(5.1,4.1,5,2.1))
  
  for (i in 1:nrow(indicators)) {
    
    indic <- indicators[i,]
    
    names <- c("f0to10E3", "f10E3to10E4", "f10E4to10E5", "f10E5to10E6", "f10E6to10E7", "f10E7")
    intensities_y <- c()
    
    for (colname in names) {
      intensities_y <- c(intensities_y, indic[colname])
    }
    intensities_y <- unlist(intensities_y)
    intensities_y <- as.numeric(levels(intensities_y))[intensities_y]
    
    title <- as.character(indic$title)
    title <- strsplit(title, "_")[[1]]
    plot_title <- ""
    for (i in 1:length(title)) {
      plot_title <- paste(plot_title, title[[i]])
      if (i %% 2 == 0)
        plot_title <- paste(plot_title, "\n")
      else
        plot_title <- paste(plot_title, " ")
    }
    
    barplot(intensities_y, names.arg = names, xlab = "intensity", ylab = "number of ions",
            main = plot_title, cex.main = 0.8)
  
  }
  
  dev.off()
  
}


graph_results <- function(indicators, criteria) {
  # Plots the results.
  #
  # Args:
  #   indicators: dataframe (one row per test) containing the results.
  
  first_criteria <- grep("nb_pseudospectra", names(indicators))
  nb_params <- first_criteria - 2
  this_criteria <- grep(criteria, names(indicators))
  
  # values taken by each parameter
  values <- list()
  for (i in 2:(1 + nb_params)) {
    lev <- unique(indicators[,i])
    values[[i - 1]] <- lev
  }
  
  length <- lapply(values, length)
  
  # indexes of the parameters taking the most values
  longest_1 <- which.max(length); length[longest_1] <- -1
  longest_2 <- which.max(length); length[longest_2] <- -1
  longest_3 <- which.max(length); length[longest_3] <- -1
  
  # indexes of the other parameters
  shortest <- which(length != -1)
  
  # all combinations of the values taken by these parameters
  combinations <- list()
  for (s in shortest) {
    page <- indicators[,s + 1]
    combinations[[length(combinations) + 1]] <- sort(as.numeric(as.character(unique(page))))
  }
  names(combinations) <- names(indicators)[shortest + 1]
  combinations <- expand.grid(combinations)
  
  # save the plots in a pdf file
  pdf(compare_graph_out)
  nb_rows <- length(values[longest_3][[1]])
  nb_cols <- length(values[longest_2][[1]])
  par(mfrow = c(nb_rows, nb_cols), mar = c(4.5,4.5,5,1))
  
  # plot parameters
  x <- sort(as.numeric(as.character(unique(indicators[,longest_1 + 1]))))
  min_zeros <- min(as.numeric(as.character(indicators$zeros_per_line)))
  max_zeros <- max(as.numeric(as.character(indicators$zeros_per_line)))
  
  for (rowi in 1:nrow(combinations)) {
    row <- combinations[rowi,]
    title <- ""
    lines <- indicators
    
    for (coli in 1:length(row)) {
      value <- as.numeric(as.character(row[coli]))
      title <- paste(title, names(row[coli]), "=", value)
      if (coli %% 2 == 0)
        title <- paste(title, "\n")
      else
        title <- paste(title, " ")
      lines <- lines[lines[names(row[coli])] == value,]
    }
    
    # values taken by the parameters
    in_plot <- lines[,longest_1 + 1]
    vertical <- lines[,longest_2 + 1]
    horizontal <-  lines[,longest_3 + 1]
    
    # for each horizontal value
    for (horiz in sort(unique(horizontal))) {
      
      # for each vertical value
      for (vertic in sort(unique(vertical))) {
        
        y <- c()
        for (this_y in sort(unique(in_plot))) {
          # line <- line[page == p & horizontal == horiz & vertical == vertic & in_plot == this_y,]
          line <- lines[horizontal == horiz & vertical == vertic & in_plot == this_y,]
          if (nrow(line) != 1)
            stop("To plot the results, each set of the parameters' values must be represented exactly 1 time")
          
          value <- line[,this_criteria]
          value <- as.numeric(as.character(value))
          y <- c(y, value)
        }
        
        this_title <- paste(title,
                       names(indicators)[longest_3 + 1], "=", horiz, "\n",
                       names(indicators)[longest_2 + 1], "=", vertic)
        
        # plot this graph
        plot(x, y, ylim = c(min_zeros, max_zeros),
             type = "b",
             xlab = names(indicators)[longest_1 + 1],
             ylab = criteria,
             main = this_title,
             cex.main = 0.8)
        
      }
      
    }
  
  }
  
  par(mfrow = c(1,1))
  dev.off()
  
}


############
#   main   #
############

update_list_of_file_names()
summary <- create_summary(nb_cores, count_duplicates = count_duplicates)

# settings_list <- summary[2][[1]]
# indicators_ <- summary[1][[1]]
indicators_ <- summary

indicators <- indicators_[order(indicators_$zeros_per_line),] # sort by number of zeros per line
# indicators_ <- indicators[order(as.numeric(row.names(indicators))),] # order back to row numbers

# record the summary in a file
indicators_to_write <- indicators[, !names(indicators) %in%
                                    c("title", "f0to10E3", "f10E3to10E4", "f10E4to10E5",
                                      "f10E5to10E6", "f10E6to10E7", "f10E7")]
write.table(indicators_to_write,
            file = summary_out,
            quote = FALSE,
            row.names = FALSE,
            sep = "\t")

# ions per intensity graph
ions_per_intensity(indicators_)

# plots to compare each set of parameters
graph_results(indicators_, criteria = "zeros_per_line")

# zip of pseudospectra .msp and .tsv files
system(paste0('cd Pseudospectra ; ls . | grep -e "msp$" -e "tsv$" | zip -r -@ "peakspectra_out.zip"'))
system(paste("cd Pseudospectra ; mv peakspectra_out.zip", peakspectra_out))

# zip of count_duplicates .tsv files
if (count_duplicates) {
  system(paste0('cd count_duplicates ; ls . | grep "tsv$" | zip -r -@ "count_duplicates_out.zip"'))
  system(paste("cd count_duplicates ; mv count_duplicates_out.zip", count_duplicates_out))
}