R/validate-by-study.R

In Sage-Bionetworks/dccmonitor: Monitor Data Validation

#' @title Validate all studies
#'
#' @description Validate metadata and manifest files for every study
#' in consortium. Returns a named list of validation results, where the name
#' refers to the study name. The validation checks are done via
#' `dccvalidator::check_all()`.
#'
#' @param fileview The fileview table.
#' @param annotations A data frame of annotation definitions.
#'   Must contain at least three columns: key, value, and columnType.
#' @param syn Synapse client object.
#' @return Named list of check results, where the names of the list refer
#'   to the study. Returns NULL if no studies in fileview.
validate_all_studies <- function(fileview, annotations, syn) {

  # Get the list of study names
  study_names <- unique(fileview$study)

  if (length(study_names) == 0) {
    return(NULL)
  }
  # Run validation checks on each set of study files
  full_results <- purrr::map(
    study_names,
    function(name) {
      study_table <- filter_study_table_latest(fileview, name)
      results <- validate_study(
        study_table = study_table,
        annotations = annotations,
        syn = syn,
        study = name
      )
      results
    }
  )
  names(full_results) <- study_names
  full_results
}

#' @title Validate a study
#'
#' @description Run validation checks for a single study and return the
#' results. Validation checks are run via `dccvalidator::check_all()`. It
#' is assumed that there is, at most, one row per metadataType in the
#' `study_table` passed in.
#'
#' @importFrom dccvalidator check_all
#' @param study_table Tibble with fileview information for
#'   a single study. Expected columns are: 'metadataType',
#'   'file_data' (tibble column with data for each file), 'assay',
#'   'species', 'biospecimen_type' (optional). Should have, at most, one row
#'   per metadataType.
#' @param annotations A data frame of annotation definitions.
#'   Must contain at least three columns: 'key', 'value', and 'columnType'.
#' @param syn Synapse client object.
#' @param study Name of the study.
#' @return Named list of check results. Returns `NULL` if there are no rows.
validate_study <- function(study_table, annotations, syn, study) {

  # Check that there is one row per metadata type
  # If not, add "dummy" row
  # Required by dccvalidator::check_all()
  required <- c("manifest", "assay", "biospecimen", "individual")
  if (!any(required %in% study_table$metadataType)) {
    return(NULL)
  }
  # Check if biospecimenType exists in table
  # If yes, make sure it's a character column
  if ("biospecimenType" %in% names(study_table)) {
    class(study_table$biospecimenType) <- "character"
  }
  if (!all(required %in% study_table$metadataType)) {
    present_types <- which(required %in% study_table$metadataType)
    for (type in required[-present_types]) {
      study_table <- tibble::add_row(
        study_table,
        metadataType = type
      )

    }
  } # else all present and it's okay to move on

  # Grab all metadata templates from config
  study_table <- add_template_col(study_table = study_table)
  samples_table = get_golem_config("samples_table")
  if (is.null(samples_table)) samples_table <- NA
  results <- dccvalidator::check_all(
    data = study_table,
    annotations = annotations,
    syn = syn,
    study = study,
    samples_table = samples_table
  )
  results
}

#' @title Add template column
#'
#' @description Add a template column to the `study_table`. The `study_table`
#' requires columns metadataType, assay, species, and allows optional
#' biospecimenType. Must have one row per `metadataType`: individual, assay,
#' biospecimen, manifest.
#'
#' @noRd
#' @inheritParams validate_study
add_template_col <- function(study_table) {
  # Ensure the needed columns exist
  if (!all(c("metadataType", "assay", "species") %in% names(study_table))) {
    stop("The study table should have the columns metadataType, assay, species")
  }
  if ("biospecimenType" %in% names(study_table)) {
    biospecimen_type <- unique(stats::na.omit(study_table[["biospecimenType"]]))
    biospecimen_type[biospecimen_type == "NaN"] <- NA # Convert NaN to NA
    biospecimen_type <- unique(stats::na.omit(biospecimen_type))
    # It's possible to get a column of NaN, NA, NULL; check length in case
    if (any(c(
      length(biospecimen_type) < 1),
      biospecimen_type %in% c("NaN", ""))
    ) {
      biospecimen_type <- NA
    }
  } else {
    biospecimen_type <- NA
  }
  species <- unique(stats::na.omit(study_table[["species"]]))
  assay <- unique(stats::na.omit(study_table[["assay"]]))
  study_table[, "template"] <- unlist(purrr::map(
    study_table[["metadataType"]],
    function(x) {
      switch(x,
             manifest = dccvalidator::gather_template_ids("manifest"),
             individual = dccvalidator::gather_template_ids(
               "individual",
               species = species
             ),
             assay = {
               if (any(c(length(assay) < 1, !is.na(assay)))) {
                 dccvalidator::gather_template_ids(
                   "assay",
                   species = species,
                   assay = assay
                 )
               } else {
                 return(NA)
               }
             },
             biospecimen = {
               if (!is.na(biospecimen_type)) {
                 template <- dccvalidator::gather_template_ids(
                   "biospecimen",
                   species = species,
                   biospecimen_type = biospecimen_type
                 )
               } else {
                 template <- dccvalidator::gather_template_ids(
                   "biospecimen",
                   species = species
                 )
               }
               # Should only have a single template, but if there was no biospecimen
               # and the config has biospecimen types, then will have all possible
               # Just return NA in that case
               if (length(template) > 1) {
                 return(NA)
               }
               template
             }
      )
    }
  ))
  return(study_table)
}