get_TCGA_project_MAF.R

#' Get MAF data from TCGA cohort
#' 
#' This convenience function queries the Genomic Data Commons API to get MAF data
#' generated with the Aliquot Ensemble Somatic Variant Merging and Masking workflow for
#' the specified project, and writes an MAF file. The API always provides data from the
#' latest data release. This function might work with non-TCGA MAF data hosted on GDC
#' (e.g., TARGET and GENIE-MSK), but it hasn't been tested and users should proceed with
#' caution.
#' 
#' TCGA cohort MAFs will be structured as downloaded, with a Unique_Patient_Identifier
#' column generated from the first 12 characters of Tumor_Sample_Barcode. When passed to
#' preload_maf() or load_maf(), this column will supersede Tumor_Sample_Barcode. In the
#' handful of patients with multiple Tumor_Sample_Barcodes (essentially replicated
#' sequencing, with very high variant overlap), these functions will effectively take the
#' union of these samples for each patient. Relatedly, the small number of TCGA
#' non-primary tumor samples should not be handled this way (and such samples are by
#' default removed by this function).
#' 
#' Temporary aliquot MAF files downloaded by this function are deleted after they are read.
#' 
#' @param project TCGA project name (e.g., "TCGA-BRCA").
#' @param filename Output filename where MAF data should be saved. Must end in '.maf'
#'   (plaintext) or '.maf.gz' (gzip compressed).
#' @param exclude_TCGA_nonprimary Default TRUE. For TCGA projects, exclude samples not
#'   associated with a patient's initial primary tumor. (In many TCGA projects, a small
#'   handful of patients have metastatic, recurrent, or additional primary samples.)
#' @param test_run Default FALSE. When TRUE, gets MAF data for a few samples instead of the whole cohort.
#' @export
get_TCGA_project_MAF = function(project = NULL, filename = NULL, test_run = FALSE,
                               exclude_TCGA_nonprimary = TRUE) {
  if (! is.character(project) || length(project) != 1) {
    stop("project should be 1-length character (e.g., TCGA-BRCA).")
  }
  
  if (! is.character(filename) || length(filename) != 1) {
    stop("filename should be a pathname for the output MAF file.")
  }
  if (file.exists(filename)) {
      stop('Specified output filename already exists.')
  }
  dir = dirname(filename)
  if(! dir.exists(dir)) {
    stop("The directory specified in the output file path does not exist.")
  }
  
  if(file.access(dir, 2) != 0) {
    if (dir == '.') {
      msg = paste0("You don't have write permissions in your working directory.",
                   " Change directories or specify a different path for your output filename.")
      stop(pretty_message(msg, emit = F))
    } else {
      stop("The directory specified in the output file path is not writeable.")
    }
  }
  
  if(endsWith(filename, '.maf.gz')) {
    gzipped = TRUE
  } else if(endsWith(filename, '.maf')) {
    gzipped = FALSE
  } else {
    stop("filename must end in .maf or .maf.gz")
  }
  
  if (file.create(filename, showWarnings = FALSE)) {
    file.remove(filename)
  } else {
    stop("The specified filename couldn't be used.")
  }
  
  if(! is.logical(exclude_TCGA_nonprimary) || length(exclude_TCGA_nonprimary) != 1) {
    stop("remove_nonprimary should be T/F.")
  }
  
  if(! is.logical(test_run) || length(test_run) != 1) {
    stop("test_run should be T/F.")
  }
  
  projects_endpt = 'https://api.gdc.cancer.gov/projects'
  files_endpt = 'https://api.gdc.cancer.gov/files'
  data_endpt = 'https://api.gdc.cancer.gov/data'
  versions_endpt = 'https://api.gdc.cancer.gov/files/versions'
  
  response = httr::GET(projects_endpt, 
                       query = list(fields = "project_id", size = '10000', format = 'JSON'))
  if (response$status_code != 200) {
    msg = paste0("GDC API query (for names of valid TCGA projects) failed ", "(got response ", 
                  response$status_code, "). Perhaps there's a problem with your network;",
                 " alternatively, check the GDC site to ensure that the data portal is online.")
    stop(pretty_message(msg, emit = F))
  }
  tcga_projects = sapply(rjson::fromJSON(rawToChar(response$content))$data$hits, '[', 'project_id')
  if(! project %in% tcga_projects) {
    fixed_project = paste0('TCGA-', project)
    if (fixed_project %in% tcga_projects) {
      project = fixed_project
    } else {
      pretty_message(paste0("TCGA projects: ", paste(tcga_projects, collapse = ", "), "."))
      stop("project is not a valid TCGA project name.")
    }
  }
  
  if(project == 'TCGA-SKCM' & exclude_TCGA_nonprimary == TRUE) {
    msg = paste0("Unlike other TCGA projects, SKCM is mostly metastatic samples. If ",
                 "you want to include metastatic patients in the cohort, re-run with ",
                 "exclude_TCGA_nonprimary = FALSE.")
    warning(pretty_message(msg, emit = F), immediate. = TRUE)
  }
  
  # We'll only remove the non-primary tumors for the user
  is_tcga_project = startsWith(project, 'TCGA')
  
  if (! exclude_TCGA_nonprimary && ! is_tcga_project) {
    warning("Setting exclude_TCGA_nonprimary = FALSE doesn't do anything when the project is not a TCGA project.")
  }
  
  exclude_TCGA_nonprimary = exclude_TCGA_nonprimary && is_tcga_project
  
  user_call = paste0("# Generated with cancereffectsizeR v", packageVersion('cancereffectsizeR'),
                   ': get_TCGA_project_MAF(project = ', deparse(project), ', test_run = ', deparse(test_run))
  if(is_tcga_project) {
    user_call = paste0(user_call, ", exclude_TCGA_nonprimary = ", deparse(exclude_TCGA_nonprimary))
  }
  user_call = paste0(user_call, ')')
  
  filters = sprintf(
  '{
    "op": "and",
    "content": [
      {
        "op": "in",
        "content":{
          "field": "cases.project.project_id",
          "value": ["%s"]
        }
      },
      {
        "op": "in",
        "content":{
          "field": "analysis.workflow_type",
          "value": ["Aliquot Ensemble Somatic Variant Merging and Masking"]
        }
      },
      {
        "op": "in",
        "content":{
          "field": "access",
          "value": ["open"]
        }
      }
    ]
  }', project)
  
  num_files = ifelse(test_run, '5', '100000')
  response = httr::GET(files_endpt, 
                       query = list(filters = filters, 
                                    fields = "file_name,md5sum,release.version", size = num_files, format = 'JSON'))
  
  if (response$status_code != 200) {
    msg = paste0("Could not get list of ", project, " cases (GDC API query failed with status code ", 
         response$status_code, ")." )
    stop(pretty_message(msg, emit = F))
  }

  content = rjson::fromJSON(rawToChar(response$content))
  files = rbindlist(lapply(content$data$hits, '[', c("id", "file_name", "md5sum")))
  if(files[, .N] == 0) {
    stop("No matching TCGA participants (patients).")
  }
  
  message("Downloading ", files[, .N] , " temporary MAF files....")
  tmp_dir = tempdir()
  files[, path := paste0(tmp_dir, '/', file_name)]
  files[, url := paste0(data_endpt, '/', id)]
  
  pbapply::pbmapply(
    function(url, path, filename) {
      # Try up to 5 times to download a file
      num_tries = 5

      for(i in 1:(num_tries - 1)) {
        unlink(path) # in case file exists from previous attempt
        code = 1L
        tryCatch(
          { code = utils::download.file(url = url, destfile = path, quiet = TRUE, mode = 'wb')},
          error = function(e) NULL, warning = function(w) NULL
        )
        if(is.integer(code) && code == 0) return()
        Sys.sleep(1) # wait a second
      }
      
      # On last try, we'll let warnings/errors go through.
      message("\nHaving some trouble with a download. Waiting 10 seconds and giving it one last try....\n")
      Sys.sleep(10)
      unlink(path)
      withCallingHandlers(
        {
          code = utils::download.file(url = url, destfile = path, quiet = TRUE, mode = 'wb')
        }, error = function(e) {
          msg = pretty_message(paste0("File ", filename, " failed to download from ",
                                      url, " (tried ", num_tries, " times)."), emit = F)
          warning("\n", msg, "\nLast errors/warnings:\n", call. = FALSE, immediate. = TRUE)
          e
        }, warning = function(w) w
      )
    }, files$url, files$path, files$file_name)
  
  message("Verifying files....")
  actual_sums = tools::md5sum(files$path)
  files[names(actual_sums), obs_sum := actual_sums, on = 'path']
  files[, failed := TRUE]
  files[obs_sum == md5sum, failed := FALSE]
  
  num_failed = files[failed == TRUE, .N]
  if(num_failed > 0) {
    failing_urls = files[failed == TRUE, url]
    if(length(failing_urls) > 20) {
      failing_urls = c(failing_urls[1:15], paste0("and ", length(failing_urls) - 15, " more."))
    }
    failing_urls = paste(failing_urls, collapse = ",\n")
    unlink(files$path)
    stop("Some files did not download correctly:\n", failing_urls)
  }
  
  to_read = files$path
  names(to_read) = files$id
  cohort_maf = rbindlist(lapply(to_read, fread, skip = 'Hugo'), idcol = "source_file_id") # column headers start with Hugo_Symbol (comment lines precede)
  
  if(is_tcga_project) {
    cohort_maf[, Unique_Patient_Identifier := substr(Tumor_Sample_Barcode, 1, 12)]
    cohort_maf[, c("V1", "V2", "V3", "type_vial", "portion_analyte", "plate", "center") := tstrsplit(Tumor_Sample_Barcode, split = "-")]
    cohort_maf[, c("V1", "V2", "V3", "portion_analyte", "plate", "center") := NULL] # already extracted participant ID
    cohort_maf[, tissue_type := substr(type_vial, 1, 2)]
    
    # 01 = primary solid tumor; 03 = primary blood-derived cancer (sample of peripheral blood)
    # (from https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/sample-type-codes)
    cohort_maf = cohort_maf[, primary_sample := tissue_type %in% c("01", "03")]
    
    num_not_primary = cohort_maf[primary_sample == FALSE, uniqueN(Tumor_Sample_Barcode)]
    if(num_not_primary > 0) {
      if (exclude_TCGA_nonprimary) {
        pretty_message(paste0("Removing ", num_not_primary, " samples that are not of tissue types 01 (primary solid tumor)",
                              " or 03 (primary blood-derived cancer)."))
        cohort_maf = cohort_maf[primary_sample == TRUE]
      } else {
        pretty_message("Note: Some samples are not of tissue types 01 (primary solid tumor) or 03 (primary blood-derived cancer).")
      }
    }
  } else {
    msg = paste0("Tumor_Sample_Barcode parsing is not implemented for non-TCGA projects. Verify that no patient has multiple samples ",
                 "before assuming that variants associated with different Tumor_Sample_Barcodes are independent.")
    warning(pretty_message(msg, emit = F))
  }

  message("Deleting temporary files....")
  unlink(files$path)
  
  if(is_tcga_project) {
    num_samples = uniqueN(cohort_maf$Tumor_Sample_Barcode)
    num_participants = uniqueN(cohort_maf$Unique_Patient_Identifier)
    message("Writing MAF file covering ", num_samples , " samples from ",
                          num_participants, " patients....")
    
    
    cohort_maf[, multisample_patient := uniqueN(Tumor_Sample_Barcode) > 1, by = "Unique_Patient_Identifier"]
    
    samples_per_tissue_type = unique(cohort_maf[, .(tissue_type, Tumor_Sample_Barcode, Unique_Patient_Identifier)])
    multiple_samples_same_tissue = samples_per_tissue_type[, .N, by = c("Unique_Patient_Identifier", "tissue_type")][N > 1, Unique_Patient_Identifier]
    num_multisample_same_tissue = uniqueN(multiple_samples_same_tissue) 
    if(num_multisample_same_tissue > 0) {
      msg = paste0(num_multisample_same_tissue, " patients have multiple sequenced samples ",
                   "that ultimately derive from the same tissue sample. We typically merge somatic calls by patient for these samples. ",
                   "In cancereffectsizeR, preload_maf() will automatically merge and de-duplicate records by using the Unique_Patient_Identifier field, rather than ",
                   "Tumor_Sample_Barcode. In non-cancereffectsizeR analyses, make sure that same-patient samples are not inadvertently treated as coming ",
                   "from different patients. (These samples are marked in the multisample_patient field.)")
      pretty_message(msg)
      message()
      setcolorder(cohort_maf, 'multisample_patient')
    }
    
    cohort_maf[, multitissue_patient := uniqueN(tissue_type) > 1, by = "Unique_Patient_Identifier"]
    
    if(any(cohort_maf$multitissue_patient)) {
      msg = paste0("Some patients, marked in a multitissue_patient column, have samples from multiple tissue sources. (See ",
          "https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/sample-type-codes for tissue descriptions.) Since the phylogenetic ",
          "relationships between these samples are unclear, consider excluding these patients from cancereffectsizeR ",
          "analyses (or establish criteria for which samples to use from each patient).")
      pretty_message(msg)
      setcolorder(cohort_maf, 'multitissue_patient')
    } else {
      cohort_maf[, multitissue_patient := NULL]
    }
    setcolorder(cohort_maf, c('Unique_Patient_Identifier', 'Tumor_Sample_Barcode'))
    cohort_maf[, c("type_vial", "tissue_type", "primary_sample") := NULL]
  } else {
    num_samples = uniqueN(cohort_maf$Tumor_Sample_Barcode)
    
    message("Writing MAF file covering ", num_samples, " samples.")
  }
  
  # Get current GDC release (one would think there would be a simpler way)
  response = httr::GET(paste0(versions_endpt, '/', cohort_maf$source_file_id[1], "?format=JSON"))
  if (response$status_code != 200) {
    warning("Failed to add GDC release to MAF file header, due to a failed API call (weird!).")
    latest_release = "Unknown (at least 33.1)"
  } else {
    latest_release = tail(rjson::fromJSON(rawToChar(response$content))[[1]]$latest_release, n = 1)
  }
  
  if (gzipped) {
    out = gzfile(filename)
  } else {
    out = file(filename)
  }
  
  headers = c(user_call, 
              paste0("# MAF data from Genomic Data Commons data release ", latest_release, 
                     " (Aliquot Ensemble Somatic Variant Merging and Masking workflow)."),
              paste(names(cohort_maf), collapse = "\t"))
  writeLines(headers, out)
  close(out)
  fwrite(cohort_maf, filename, append = TRUE, sep = "\t")
  message(paste0("MAF file saved to ", filename, "."))
}