changeset 8:1574b8a29e6f draft

planemo upload for repository https://github.com/RECETOX/galaxytools/tree/master/tools/recetox_aplcms commit bc3445f7c41271b0062c7674108f57708d08dd28
author recetox
date Thu, 30 May 2024 14:52:30 +0000
parents d69ec1171d71
children 9064aecc78a7
files help.xml utils.R
diffstat 2 files changed, 79 insertions(+), 79 deletions(-) [+]
line wrap: on
line diff
--- a/help.xml	Wed Oct 11 11:18:10 2023 +0000
+++ b/help.xml	Thu May 30 14:52:30 2024 +0000
@@ -49,7 +49,7 @@
 tolerance levels are estimated from the data. A run-filter is used to detect peaks and remove noise.
 Non-parametric statistical methods are used to find-tune peak selection and grouping. After retention time
 correction, a feature table is generated by aligning peaks across spectra. For further information on apLCMS
-please refer to https://mypage.cuhk.edu.cn/academics/yutianwei/apLCMS/.
+please refer to the official tutorial.
 </token>
 
 <token name="@REMOVE_NOISE_HELP@">
--- a/utils.R	Wed Oct 11 11:18:10 2023 +0000
+++ b/utils.R	Thu May 30 14:52:30 2024 +0000
@@ -1,139 +1,139 @@
 library(recetox.aplcms)
 
 get_env_sample_name <- function() {
-  sample_name <- Sys.getenv("SAMPLE_NAME", unset = NA)
-  if (nchar(sample_name) == 0) {
-    sample_name <- NA
-  }
-  if (is.na(sample_name)) {
-    message("The mzML file does not contain run ID.")
-  }
-  return(sample_name)
+    sample_name <- Sys.getenv("SAMPLE_NAME", unset = NA)
+    if (nchar(sample_name) == 0) {
+        sample_name <- NA
+    }
+    if (is.na(sample_name)) {
+        message("The mzML file does not contain run ID.")
+    }
+    return(sample_name)
 }
 
 save_sample_name <- function(df, sample_name) {
-  attr(df, "sample_name") <- sample_name
-  return(df)
+    attr(df, "sample_name") <- sample_name
+    return(df)
 }
 
 restore_sample_name <- function(df) {
-  return(df$sample_id[1])
+    return(df$sample_id[1])
 }
 
 load_sample_name <- function(df) {
-  sample_name <- attr(df, "sample_name")
-  if (is.null(sample_name)) {
-    return(NA)
-  } else {
-    return(sample_name)
-  }
+    sample_name <- attr(df, "sample_name")
+    if (is.null(sample_name)) {
+        return(NA)
+    } else {
+        return(sample_name)
+    }
 }
 
 save_data_as_parquet_file <- function(data, filename) {
-  arrow::write_parquet(data, filename)
+    arrow::write_parquet(data, filename)
 }
 
 load_data_from_parquet_file <- function(filename) {
-  return(arrow::read_parquet(filename))
+    return(arrow::read_parquet(filename))
 }
 
 load_parquet_collection <- function(files) {
-  features <- lapply(files, arrow::read_parquet)
-  features <- lapply(features, tibble::as_tibble)
-  return(features)
+    features <- lapply(files, arrow::read_parquet)
+    features <- lapply(features, tibble::as_tibble)
+    return(features)
 }
 
 save_parquet_collection <- function(feature_tables, sample_names, subdir) {
-  dir.create(subdir)
-  for (i in seq_len(length(feature_tables))) {
-    filename <- file.path(subdir, paste0(sample_names[i], ".parquet"))
-    feature_table <- as.data.frame(feature_tables[[i]])
-    feature_table <- save_sample_name(feature_table, sample_names[i])
-    arrow::write_parquet(feature_table, filename)
-  }
+    dir.create(subdir)
+    for (i in seq_len(length(feature_tables))) {
+        filename <- file.path(subdir, paste0(sample_names[i], ".parquet"))
+        feature_table <- as.data.frame(feature_tables[[i]])
+        feature_table <- save_sample_name(feature_table, sample_names[i])
+        arrow::write_parquet(feature_table, filename)
+    }
 }
 
 sort_by_sample_name <- function(tables, sample_names) {
-  return(tables[order(sample_names)])
+    return(tables[order(sample_names)])
 }
 
 save_tolerances <- function(table, tol_file) {
-  mz_tolerance <- c(table$mz_tol_relative)
-  rt_tolerance <- c(table$rt_tol_relative)
-  arrow::write_parquet(data.frame(mz_tolerance, rt_tolerance), tol_file)
+    mz_tolerance <- c(table$mz_tol_relative)
+    rt_tolerance <- c(table$rt_tol_relative)
+    arrow::write_parquet(data.frame(mz_tolerance, rt_tolerance), tol_file)
 }
 
 save_aligned_features <- function(aligned_features, metadata_file, rt_file, intensity_file) {
-  save_data_as_parquet_file(aligned_features$metadata, metadata_file)
-  save_data_as_parquet_file(aligned_features$rt, rt_file)
-  save_data_as_parquet_file(aligned_features$intensity, intensity_file)
+    save_data_as_parquet_file(aligned_features$metadata, metadata_file)
+    save_data_as_parquet_file(aligned_features$rt, rt_file)
+    save_data_as_parquet_file(aligned_features$intensity, intensity_file)
 }
 
 select_table_with_sample_name <- function(tables, sample_name) {
-  sample_names <- lapply(tables, load_sample_name)
-  index <- which(sample_names == sample_name)
-  if (length(index) > 0) {
-    return(tables[[index]])
-  } else {
-    stop(sprintf(
-      "Mismatch - sample name '%s' not present in %s",
-      sample_name, paste(sample_names, collapse = ", ")
-    ))
-  }
+    sample_names <- lapply(tables, load_sample_name)
+    index <- which(sample_names == sample_name)
+    if (length(index) > 0) {
+        return(tables[[index]])
+    } else {
+        stop(sprintf(
+            "Mismatch - sample name '%s' not present in %s",
+            sample_name, paste(sample_names, collapse = ", ")
+        ))
+    }
 }
 
 select_adjusted <- function(recovered_features) {
-  return(recovered_features$adjusted_features)
+    return(recovered_features$adjusted_features)
 }
 
 known_table_columns <- function() {
-  c(
-    "chemical_formula", "HMDB_ID", "KEGG_compound_ID", "mass", "ion.type",
-    "m.z", "Number_profiles_processed", "Percent_found", "mz_min", "mz_max",
-    "RT_mean", "RT_sd", "RT_min", "RT_max", "int_mean(log)", "int_sd(log)",
-    "int_min(log)", "int_max(log)"
-  )
+    c(
+        "chemical_formula", "HMDB_ID", "KEGG_compound_ID", "mass", "ion.type",
+        "m.z", "Number_profiles_processed", "Percent_found", "mz_min", "mz_max",
+        "RT_mean", "RT_sd", "RT_min", "RT_max", "int_mean(log)", "int_sd(log)",
+        "int_min(log)", "int_max(log)"
+    )
 }
 
 save_known_table <- function(table, filename) {
-  columns <- known_table_columns()
-  arrow::write_parquet(table$known_table[columns], filename)
+    columns <- known_table_columns()
+    arrow::write_parquet(table$known_table[columns], filename)
 }
 
 read_known_table <- function(filename) {
-  arrow::read_parquet(filename, col_select = known_table_columns())
+    arrow::read_parquet(filename, col_select = known_table_columns())
 }
 
 save_pairing <- function(table, filename) {
-  df <- table$pairing %>%
-    as_tibble() %>%
-    setNames(c("new", "old"))
-  arrow::write_parquet(df, filename)
+    df <- table$pairing %>%
+        as_tibble() %>%
+        setNames(c("new", "old"))
+    arrow::write_parquet(df, filename)
 }
 
 join_tables_to_list <- function(metadata, rt_table, intensity_table) {
-  features <- new("list")
-  features$metadata <- metadata
-  features$intensity <- intensity_table
-  features$rt <- rt_table
-  return(features)
+    features <- new("list")
+    features$metadata <- metadata
+    features$intensity <- intensity_table
+    features$rt <- rt_table
+    return(features)
 }
 
 validate_sample_names <- function(sample_names) {
-  if ((any(is.na(sample_names))) || (length(unique(sample_names)) != length(sample_names))) {
-    stop(sprintf(
-      "Sample names absent or not unique - provided sample names: %s",
-      paste(sample_names, collapse = ", ")
-    ))
-  }
+    if ((any(is.na(sample_names))) || (length(unique(sample_names)) != length(sample_names))) {
+        stop(sprintf(
+            "Sample names absent or not unique - provided sample names: %s",
+            paste(sample_names, collapse = ", ")
+        ))
+    }
 }
 
 determine_sigma_ratios <- function(sigma_ratio_lim_min = NA, sigma_ratio_lim_max = NA) {
-  if (is.na(sigma_ratio_lim_min)) {
-    sigma_ratio_lim_min <- 0
-  }
-  if (is.na(sigma_ratio_lim_max)) {
-    sigma_ratio_lim_max <- Inf
-  }
-  return(c(sigma_ratio_lim_min, sigma_ratio_lim_max))
+    if (is.na(sigma_ratio_lim_min)) {
+        sigma_ratio_lim_min <- 0
+    }
+    if (is.na(sigma_ratio_lim_max)) {
+        sigma_ratio_lim_max <- Inf
+    }
+    return(c(sigma_ratio_lim_min, sigma_ratio_lim_max))
 }