data-raw/download-clinical.R

In Syksy/curatedPCaData: Curated Prostate Cancer Data

# Packages used for processing data
library(magrittr)
library(dplyr)
library(MultiAssayExperiment)

## Prostate cancer clinical fields latest PRAD template ideally filled for all studies
#
# Store template_prad.csv as a data frame inside the package instead of a hidden .csv file
template_prad <- read.csv("data-raw/template_prad.csv", as.is = TRUE)
usethis::use_data(template_prad, overwrite = TRUE)

###
#
# Processing clinical metadata for datasets
#
###
#
## Contains:
#
# - Abida et al.
# - Baca et al.
# - Barbieri et al.
# - Barwick et al.
# - Chandran et al.
# - Friedrich et al.
# - Hieronymus et al.
# - ICGC (national subsets)
# - IGC
# - Kim et al.
# - Kunderfranco et al.
# - Ren et al.
# - Sun et al.
# - Taylor et al. (MSKCC)
# - TCGA
# - True et al.
# - Wallace et al.
# - Wang et al.
# - Weiner et al.

###
#
# Abida et al. 2019
# using cgdsr + cBioPortalData
#
###

mycgds <- cgdsr::CGDS("http://www.cbioportal.org/")
uncurated <- cgdsr::getClinicalData(mycgds, caseList = "prad_su2c_2019_cnaseq")

# cBioPortalData's metadata pull is missing some fields available only via cgdsr
mae <- cBioPortalData::cBioDataPack("prad_su2c_2019", ask = FALSE)
uncurated2 <- MultiAssayExperiment::colData(mae)
uncurated2 <- as.data.frame(uncurated2)
uncurated2$SAMPLE_ID <- gsub("-", ".", uncurated2$SAMPLE_ID)

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)
curated2 <- initial_curated_internal(
  df_rownames = uncurated2$SAMPLE_ID
)
# Order both datasets to same order
uncurated <- uncurated[match(uncurated2$SAMPLE_ID, rownames(uncurated)), ]

# cgdsr part
curated <- curated %>%
  dplyr::mutate(study_name = "Abida et al.") %>%
  # Take patient IDs from cBioPortalData pull
  dplyr::mutate(alt_sample_name = uncurated2$OTHER_SAMPLE_ID) %>%
  dplyr::mutate(patient_id = uncurated2$PATIENT_ID) %>%
  # Samples were paired for analyses
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_name = rownames(uncurated)) %>%
  dplyr::mutate(age_at_initial_diagnosis = floor(uncurated$AGE_AT_DIAGNOSIS)) %>%
  # dplyr::mutate(age_at_procurement = floor(uncurated$AGE_AT_PROCUREMENT)) %>%
  # All samples are metastatic
  dplyr::mutate(metastasis_occurrence_status = 1) %>%
  dplyr::mutate(sample_type = "metastatic") %>%
  dplyr::mutate(metastatic_site = dplyr::case_when(
    uncurated$TISSUE_SITE == "Adrenal" ~ "adrenal_gland",
    uncurated$TISSUE_SITE == "Bone" ~ "bone",
    uncurated$TISSUE_SITE == "Brain" ~ "brain",
    uncurated$TISSUE_SITE == "Liver" ~ "liver",
    uncurated$TISSUE_SITE == "LN" ~ "lymph_node",
    uncurated$TISSUE_SITE == "Lung" ~ "lung",
    uncurated$TISSUE_SITE == "Other Soft tissue" ~ "soft_tissue",
    uncurated$TISSUE_SITE == "Prostate" ~ "prostate",
    uncurated$TISSUE_SITE == "Unknown" ~ "other"
  )) %>%
  dplyr::mutate(psa = uncurated$PSA) %>%
  dplyr::mutate(gleason_grade = as.numeric(uncurated$GLEASON_SCORE)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    uncurated$GLEASON_SCORE <= 6 ~ "<=6",
    uncurated$GLEASON_SCORE == 7 ~ "7",
    uncurated$GLEASON_SCORE >= 8 ~ ">=8"
  )) %>%
  # dplyr::mutate(ar_score=uncurated$AR_SCORE) %>%
  dplyr::mutate(AR_activity = uncurated$AR_SCORE) %>%
  dplyr::mutate(genome_altered = uncurated$FRACTION_GENOME_ALTERED) %>%
  # dplyr::mutate(ABI_ENZA_exposure_status=uncurated$ABI_ENZA_EXPOSURE_STATUS) %>%
  ## Fusions were determined from gene expression
  dplyr::mutate(ERG_fusion_GEX = dplyr::case_when(
    uncurated$ETS_FUSION_DETAILS == "TMPRSS2-ERG" ~ 1,
    uncurated$ETS_FUSION_DETAILS != "TMPRSS2-ERG" ~ 0
  )) %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    uncurated$OS_STATUS == "1:DECEASED" ~ 1,
    uncurated$OS_STATUS == "0:LIVING" ~ 0
  )) %>%
  dplyr::mutate(days_to_overall_survival = as.numeric(uncurated$OS_MONTHS) * 30.5) %>%
  dplyr::mutate(other_treatment = uncurated$CHEMO_REGIMEN_CATEGORY) %>%
  dplyr::mutate(batch = uncurated$TISSUE_SOURCE_SITE)

# Other sample features collapsed
curated[, "other_feature"] <- apply(cbind(
  paste0("NEPC_SCORE=", uncurated$NEPC_SCORE),
  paste0("ETS_FUSION_DETAILS=", uncurated$ETS_FUSION_DETAILS),
  paste0("RAF1_BRAF_STATUS=", uncurated$RAF1_BRAF_STATUS),
  paste0("TMB_NONSYNONYMOUS=", uncurated$TMB_NONSYNONYMOUS),
  paste0("NEUROENDOCRINE_FEATURES=", uncurated$NEUROENDOCRINE_FEATURES),
  paste0("PATHOLOGY_CLASSIFICATION=", uncurated$PATHOLOGY_CLASSIFICATION)
), MARGIN = 1, FUN = function(x) {
  paste0(x, collapse = "|")
})
# Other patient features collapsed
curated[, "other_patient"] <- apply(cbind(
  paste0("TAXANE_EXPOSURE_STATUS=", uncurated$TAXANE_EXPOSURE_STATUS),
  paste0("AGE_AT_PROCUREMENT=", uncurated$AGE_AT_PROCUREMENT),
  paste0("CHEMO_REGIMEN_CATEGORY=", uncurated$CHEMO_REGIMEN_CATEGORY)
), MARGIN = 1, FUN = function(x) {
  paste0(x, collapse = "|")
})

clinical_abida <- curated

save(clinical_abida, file = "data-raw/clinical_abida.RData")

###
#
# Baca et al.
# Source: cBioPortalData
#
###

mae <- cBioPortalData::cBioDataPack("prad_broad_2013", ask = FALSE)
uncurated1 <- MultiAssayExperiment::colData(mae)
uncurated1 <- as.data.frame(uncurated1)

# mycgds <- cgdsr::CGDS("http://www.cbioportal.org/")
# uncurated2 <- cgdsr::getClinicalData(mycgds, caseList="prad_broad_2013_all")
# mycaselist_baca = cgdsr::getCaseLists(mycgds,"prad_broad_2013")

# create the curated object
curated <- initial_curated_internal(
  df_rownames = gsub("-", ".", rownames(uncurated1))
)

curated <- curated %>%
  # Portion from cBioPortalData-package
  dplyr::mutate(study_name = "Baca et al.") %>%
  dplyr::mutate(sample_name = gsub("-", ".", uncurated1$SAMPLE_ID)) %>%
  dplyr::mutate(patient_id = gsub("-", ".", rownames(uncurated1))) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated1$AGE) %>%
  ## TDL: Should follow allowed values assigned in 'template_prad.csv'
  # dplyr::mutate(sample_type=uncurated$SAMPLE_TYPE)%>%
  dplyr::mutate(psa = uncurated1$SERUM_PSA) %>%
  dplyr::mutate(gleason_major = as.integer(stringr::str_sub(uncurated1$GLEASON_SCORE, 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.integer(stringr::str_sub(uncurated1$GLEASON_SCORE, 3, 3))) %>%
  dplyr::mutate(T_pathological = readr::parse_number(uncurated1$PATH_T_STAGE)) %>%
  dplyr::mutate(T_substage_pathological = dplyr::case_when(
    uncurated1$PATH_T_STAGE %in% c("Metastatic") ~ "NA",
    TRUE ~ stringr::str_extract(uncurated1$PATH_T_STAGE, "[a-c]+")
  )) %>%
  dplyr::mutate(gleason_grade = gleason_major + gleason_minor) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    # Some major/minor combinations were missing while overall sum was available
    stringr::str_sub(uncurated1$GLEASON_SCORE, 1, 3) == "3+3" ~ "<=6",
    stringr::str_sub(uncurated1$GLEASON_SCORE, 1, 3) == "3+4" ~ "3+4",
    stringr::str_sub(uncurated1$GLEASON_SCORE, 1, 3) == "4+3" ~ "4+3",
    stringr::str_sub(uncurated1$GLEASON_SCORE, 1, 3) %in% c("4+4", "4+5") ~ ">=8",
    gleason_grade >= 8 ~ ">=8"
  )) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated1$SAMPLE_TYPE == "Metastasis" ~ "metastatic",
    uncurated1$SAMPLE_TYPE == "Primary" ~ "primary"
  )) %>%
  # The samples were analyzed in a paired manner to normal tissue
  dplyr::mutate(sample_paired = 1) %>%
  # ABSOLUTE purity scores
  dplyr::mutate(tumor_purity_absolute = uncurated1$PURITY_ABSOLUTE) %>%
  # Batches inside the study
  dplyr::mutate(batch = uncurated1$COHORT) %>%
  # Nonsynonymous TMB added as other sample information
  dplyr::mutate(other_sample = paste0("TMB_NONSYNONYMOUS=", round(uncurated1$TMB_NONSYNONYMOUS, 3))) %>%
  # ERG statuses from FISH or sequencing
  dplyr::mutate(ERG_fusion_IHC = as.numeric(!is.na(uncurated1$ERG_FISH_RESULT))) %>%
  # Contains also non-ERG alterations from sequencing
  dplyr::mutate(ERG_fusion_CNA = 0)

curated[grep("TMPRSS2-ERG", uncurated1$"ETS_FUSION_SEQ"), "ERG_fusion_CNA"] <- 1
# rownames(curated) <- curated$patient_id

clinical_baca <- curated

save(clinical_baca, file = "data-raw/clinical_baca.RData")

###
#
# Barbieri et al.
# cBioportal Barbieri Broad/Cornell Data
# using cBioPortalData
#
###

mae <- cBioPortalData::cBioDataPack("prad_broad", ask = FALSE)
uncurated <- colData(mae)
uncurated <- as.data.frame(uncurated)
rownames(uncurated) <- gsub("-", ".", rownames(uncurated))
# mycgds <- cgdsr::CGDS("http://www.cbioportal.org/")
# uncurated <- cgdsr::getClinicalData(mycgds, caseList="prad_broad_sequenced")

# create the curated object
curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Barbieri et al.") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(patient_id = row.names(uncurated)) %>%
  # from the publication: Primary treatment naive radical prostatectomy specimen from American and Australian patients
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(race = "caucasian") %>%
  dplyr::mutate(tissue_source = "prostatectomy") %>%
  dplyr::mutate(gleason_major = as.integer(stringr::str_sub(uncurated$GLEASON_SCORE, 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.integer(stringr::str_sub(uncurated$GLEASON_SCORE, 3, 3))) %>%
  dplyr::mutate(gleason_grade = gleason_major + gleason_minor) %>%
  # dplyr::mutate(gleason_grade = gsub(";.*","",uncurated$GLEASON_SCORE)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    stringr::str_sub(uncurated$GLEASON_SCORE, 1, 3) == "3+3" ~ "<=6",
    stringr::str_sub(uncurated$GLEASON_SCORE, 1, 3) == "3+4" ~ "3+4",
    stringr::str_sub(uncurated$GLEASON_SCORE, 1, 3) == "4+3" ~ "4+3",
    stringr::str_sub(uncurated$GLEASON_SCORE, 1, 3) %in% c("4+4", "4+5") ~ ">=8"
  )) %>%
  dplyr::mutate(psa = uncurated$SERUM_PSA) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$AGE) %>%
  dplyr::mutate(T_clinical = readr::parse_number(uncurated$TUMOR_STAGE)) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(uncurated$TUMOR_STAGE, "[a-c]+")) %>%
  # from the publication: TMPRSS2-ERG fusion status assessed by fluorescence in situ hybridization (FISH)
  # dplyr::mutate(TMPRSS2_ERG_FUSION_STATUS=uncurated$TMPRSS2_ERG_FUSION_STATUS) %>%
  dplyr::mutate(ERG_fusion_IHC = dplyr::case_when(
    uncurated$TMPRSS2_ERG_FUSION_STATUS == "Negative" ~ 0, # Negative -> 0
    TRUE ~ 1 # Else -> 1
  )) %>%
  dplyr::mutate(GLEASON_SCORE_PERCENT_4_AND_5 = uncurated$GLEASON_SCORE_PERCENT_4_AND_5)

clinical_barbieri <- curated

save(clinical_barbieri, file = "data-raw/clinical_barbieri.RData")

####
#
# Barwick et al.
# curated from GEO's clinical metadata (GSM)
#
####

gse <- GEOquery::getGEO("GSE18655", GSEMatrix = TRUE)

uncurated <- Biobase::pData(gse[[1]])

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Barwick et al.") %>%
  dplyr::mutate(patient_id = paste0("X", uncurated$"title")) %>%
  dplyr::mutate(sample_name = paste0("X", uncurated$"title")) %>%
  dplyr::mutate(age_at_initial_diagnosis = as.numeric(uncurated$"age:ch1")) %>%
  dplyr::mutate(psa = as.numeric(uncurated$"psa:ch1")) %>%
  dplyr::mutate(gleason_grade = as.numeric(uncurated$"gleason score:ch1")) %>%
  dplyr::mutate(days_to_disease_specific_recurrence = round(30.5 * as.numeric(uncurated$"follow-up (months):ch1"), 0)) %>%
  dplyr::mutate(disease_specific_recurrence_status = as.numeric(uncurated$"recurrence:ch1" == "Rec")) %>%
  dplyr::mutate(tumor_margins_positive = as.numeric(uncurated$"positive surgical margin:ch1" == "Positive Margin")) %>%
  dplyr::mutate(tissue_source = "prostatectomy") %>%
  # Based on the publication text, ERG-fusion status was determined using over-expression of ERG-transcripts in gene expression and then experimentally validated
  dplyr::mutate(ERG_fusion_GEX = as.numeric(uncurated$"tmprss2:ch1" == "ERG Fusion: Fusion")) %>%
  # It appears the T staging was based on pathology
  dplyr::mutate(T_pathological = as.numeric(uncurated$"grade:ch1")) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(sample_paired = 0)

# Replicates provided in the raw gene expression matrix:
#> grep("rep", colnames(gex), value=TRUE)
# [1] "X1_rep1"   "X1_rep2"   "X100_rep1" "X100_rep2" "X105_rep1" "X105_rep2" "X151_rep1" "X151_rep2" "X172_rep1" "X172_rep2" "X61_rep1"  "X61_rep2"  "X77_rep1"  "X77_rep2"  "X9_rep1"   "X9_rep2"
#> unique(unlist(lapply(grep("rep", colnames(gex), value=TRUE), FUN=function(x) { strsplit(x, "_")[[1]][1]})))
# [1] "X1"   "X100" "X105" "X151" "X172" "X61"  "X77"  "X9"

# Append correct "_rep1", "_rep2" suffixes to correct samples
curated <- rbind(curated, curated[which(curated$patient_id %in% c("X1", "X100", "X105", "X151", "X172", "X61", "X77", "X9")), ])
curated[curated$sample_name %in% c("X1", "X100", "X105", "X151", "X172", "X61", "X77", "X9"), "sample_name"] <- paste0(curated$sample_name[curated$sample_name %in% c("X1", "X100", "X105", "X151", "X172", "X61", "X77", "X9")], rep(c("_rep1", "_rep2"), each = 7))
rownames(curated) <- NULL

clinical_barwick <- curated

save(clinical_barwick, file = "./data-raw/clinical_barwick.RData")

###
#
# Chandran et al., BMC Cancer 2007
# Yu et al. J Clin Oncol 2004
#
# https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE6919
#
# Includes normal samples, tumor samples, as well as metastatic samples
# using GEOquery's sample fields (GSM)
#
###

gse <- GEOquery::getGEO("GSE6919", GSEMatrix = TRUE)

uncurated1 <- Biobase::pData(gse[[1]]) # Batch 1
uncurated2 <- Biobase::pData(gse[[2]]) # Batch 2
uncurated3 <- Biobase::pData(gse[[3]]) # Batch 3
# Bind metadata over the 3 GPLs
uncurated <- rbind(uncurated1, uncurated2, uncurated3)

# Base curation metadat
curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

# Pipe through the available fields
curated <- curated %>%
  dplyr::mutate(study_name = "Chandran et al.") %>%
  dplyr::mutate(patient_id = uncurated$"title") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  # dplyr::mutate(alt_sample_name = uncurated$'title') %>%
  dplyr::mutate(age_at_initial_diagnosis = as.numeric(uncurated$"Age:ch1")) %>%
  dplyr::mutate(race = dplyr::case_when(
    uncurated$"Race:ch1" == "Caucasian" ~ "caucasian",
    uncurated$"Race:ch1" == "African American" ~ "african_american"
  )) %>%
  dplyr::mutate(gleason_grade = uncurated$"Gleason Grade:ch1") %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    uncurated$"Gleason Grade:ch1" %in% 4:6 ~ "<=6",
    uncurated$"Gleason Grade:ch1" == 7 ~ "7",
    uncurated$"Gleason Grade:ch1" %in% 8:10 ~ ">=8"
  )) %>%
  # T_pathological & T_substage_pathological OR T_clinical & T_substage_clinical
  dplyr::mutate(T_pathological = dplyr::case_when(
    uncurated$"Tumor stage:ch1" %in% c("T2a", "T2b") ~ 2,
    uncurated$"Tumor stage:ch1" %in% c("T3a", "T3b") ~ 3,
    uncurated$"Tumor stage:ch1" %in% c("T4", "T4a") ~ 4,
    is.na(uncurated$"Tumor stage:ch1") ~ NA_real_,
  )) %>%
  dplyr::mutate(T_substage_pathological = dplyr::case_when(
    uncurated$"Tumor stage:ch1" %in% c("T2a", "T3a", "T4a") ~ "a",
    uncurated$"Tumor stage:ch1" %in% c("T2b", "T3b") ~ "b",
    uncurated$"Tumor stage:ch1" == "T4" ~ NA_character_,
    is.na(uncurated$"Tumor stage:ch1") ~ NA_character_,
  )) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated$"Tissue:ch1" %in% c("primary prostate tumor") ~ "primary",
    uncurated$"Tissue:ch1" %in% c("normal prostate tissue adjacent to tumor") ~ "normal",
    uncurated$"Tissue:ch1" %in% c("normal prostate tissue free of any pathological alteration from brain-dead organ donor") ~ "normal",
    uncurated$"Tissue:ch1" %in% c("metastases recurrent in prostate", "prostate tumor metastases in adrenal gland", "prostate tumor metastases in kidney", "prostate tumor metastases in left inguinal lymph node", "prostate tumor metastases in liver", "prostate tumor metastases in lung", "prostate tumor metastases in para aortic lymph node", "prostate tumor metastases in para tracheal lymph node", "prostate tumor metastases in paratracheal lymph node", "prostate tumor metastases in retroperitoneal lymph node") ~ "metastatic",
    is.na(uncurated$"Tissue:ch1") ~ NA_character_,
  )) %>%
  dplyr::mutate(metastatic_site = dplyr::case_when(
    uncurated$"Tissue:ch1" %in% c("metastases recurrent in prostate") ~ "prostate",
    uncurated$"Tissue:ch1" %in% c("prostate tumor metastases in adrenal gland") ~ "adrenal_gland",
    uncurated$"Tissue:ch1" %in% c("prostate tumor metastases in kidney") ~ "kidney",
    uncurated$"Tissue:ch1" %in% c("prostate tumor metastases in liver") ~ "liver",
    uncurated$"Tissue:ch1" %in% c("prostate tumor metastases in lung") ~ "lung",
    uncurated$"Tissue:ch1" %in% c("prostate tumor metastases in left inguinal lymph node", "prostate tumor metastases in para aortic lymph node", "prostate tumor metastases in para tracheal lymph node", "prostate tumor metastases in paratracheal lymph node", "prostate tumor metastases in retroperitoneal lymph node") ~ "lymph_node",
    uncurated$"Tissue:ch1" %in% c("primary prostate tumor", "normal prostate tissue adjacent to tumor", "normal prostate tissue free of any pathological alteration from brain-dead organ donor") ~ NA_character_,
    is.na(uncurated$"Tissue:ch1") ~ NA_character_,
  ))

rownames(curated) <- curated$sample_name
clinical_chandran <- curated

save(clinical_chandran, file = "data-raw/clinical_chandran.RData")

###
#
# Friedrich et at German samples.
# using GEOquery's clinical data fields (GSM)
#
###

gset <- GEOquery::getGEO("GSE134051", GSEMatrix = TRUE, getGPL = TRUE)

uncurated <- Biobase::pData(gset[[1]])
curated <- initial_curated_internal(df_rownames = rownames(uncurated))

curated <- curated %>%
  dplyr::mutate(study_name = "Friedrich et al.") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(patient_id = row.names(uncurated)) %>%
  dplyr::mutate(alt_sample_name = unlist(lapply(stringr::str_split(uncurated$title, "_"), function(x) x[3]))) %>%
  dplyr::mutate(gleason_grade = dplyr::case_when(
    uncurated$"gleason score:ch1" == "None" ~ NA_character_,
    TRUE ~ uncurated$"gleason score:ch1"
  )) %>%
  dplyr::mutate(gleason_grade = as.numeric(gleason_grade)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade <= 6 ~ "<=6",
    gleason_grade == 7 ~ "7",
    gleason_grade >= 8 ~ ">=8"
  )) %>%
  dplyr::mutate(race = "caucasian") %>%
  dplyr::mutate(tissue_source = "prostatectomy") %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated$"risk group:ch1" == "C" ~ "BPH",
    uncurated$"risk group:ch1" == "V" ~ "primary",
    uncurated$"risk group:ch1" == "Ms" ~ "primary",
    uncurated$"risk group:ch1" == "Md" ~ "primary",
    uncurated$"risk group:ch1" == "L" ~ "primary",
    uncurated$"risk group:ch1" == "H+st" ~ "primary",
    uncurated$"risk group:ch1" == "H+dt" ~ "primary",
    uncurated$"risk group:ch1" == "H-st" ~ "primary",
    uncurated$"risk group:ch1" == "H-dt" ~ "primary",
    uncurated$"risk group:ch1" == "H+sf" ~ "normal",
    uncurated$"risk group:ch1" == "H+df" ~ "normal",
    uncurated$"risk group:ch1" == "H-sf" ~ "normal",
    uncurated$"risk group:ch1" == "H-df" ~ "normal"
  )) %>%
  dplyr::mutate(days_to_overall_survival = ceiling(as.numeric(uncurated$"follow-up time in months:ch1") * 30.5)) %>%
  dplyr::mutate(frozen_ffpe = "frozen") %>%
  dplyr::mutate(microdissected = 1) %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    uncurated$"risk group:ch1" == "C" ~ 0,
    uncurated$"risk group:ch1" == "V" ~ 0,
    uncurated$"risk group:ch1" == "Ms" ~ 0,
    uncurated$"risk group:ch1" == "Md" ~ 1,
    uncurated$"risk group:ch1" == "L" ~ 0,
    uncurated$"risk group:ch1" == "H+st" ~ 0,
    uncurated$"risk group:ch1" == "H+dt" ~ 1,
    uncurated$"risk group:ch1" == "H-st" ~ 0,
    uncurated$"risk group:ch1" == "H-dt" ~ 1,
    uncurated$"risk group:ch1" == "H+sf" ~ 0,
    uncurated$"risk group:ch1" == "H+df" ~ 0,
    uncurated$"risk group:ch1" == "H-sf" ~ 0,
    uncurated$"risk group:ch1" == "H-df" ~ 0
  ))

clinical_friedrich <- curated

save(clinical_friedrich, file = "data-raw/clinical_friedrich.RData")

###
#
# Hieronymus et al.
# Data pulled from GEO's clinical metadata (GSM)
#
###

gse <- GEOquery::getGEO("GSE54691", GSEMatrix = TRUE)

uncurated <- Biobase::pData(gse[[1]])

curated <- initial_curated_internal(
  df_rownames = sub(".*\\s", "", uncurated$title)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Hieronymus, et al.") %>%
  dplyr::mutate(sample_name = sub(".*\\s", "", uncurated$title)) %>%
  dplyr::mutate(patient_id = rownames(uncurated)) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    is.na(uncurated$`survivalevent:ch1`) ~ 0,
    TRUE ~ 1
  )) %>%
  dplyr::mutate(
    days_to_overall_survival =
      round(as.numeric(uncurated$`survival_or_followup_time_months:ch1`) * 30.5, 0)
  ) %>%
  dplyr::mutate(
    age_at_initial_diagnosis =
      floor(as.numeric(uncurated$`dxage:ch1`))
  ) %>%
  dplyr::mutate(gleason_grade = as.numeric(uncurated$`pathggs:ch1`)) %>%
  dplyr::mutate(gleason_minor = as.numeric(uncurated$`pathgg2:ch1`)) %>%
  dplyr::mutate(gleason_major = as.numeric(uncurated$`pathgg1:ch1`)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade == 6 ~ "<=6",
    gleason_grade >= 8 ~ ">=8",
    gleason_grade == 7 & gleason_major == "3" ~ "3+4",
    gleason_grade == 7 & gleason_major == "4" ~ "4+3",
  )) %>%
  dplyr::mutate(source_of_gleason = "prostatectomy") %>%
  dplyr::mutate(T_pathological = readr::parse_number(uncurated$`pathstage:ch1`)) %>%
  dplyr::mutate(T_substage_pathological = stringr::str_extract(
    uncurated$`pathstage:ch1`,
    "[a-c]+"
  )) %>%
  dplyr::mutate(T_clinical = readr::parse_number(uncurated$`clint_stage:ch1`)) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(
    uncurated$`clint_stage:ch1`,
    "[a-c]+"
  )) %>%
  dplyr::mutate(metastasis_occurrence_status = dplyr::case_when(
    uncurated$`metsevent:ch1` == "no" ~ 0,
    uncurated$`metsevent:ch1` == "yes" ~ 1
  )) %>%
  dplyr::mutate(days_to_metastatic_occurrence = round(as.numeric(
    uncurated$`metsfreetime_months:ch1`
  ) * 30.5, 0)) %>%
  dplyr::mutate(psa = as.numeric(uncurated$`pretxpsa:ch1`)) %>%
  dplyr::mutate(extraprostatic_extension = dplyr::case_when(
    uncurated$`ece_binary:ch1` == "No" ~ 0,
    uncurated$`ece_binary:ch1` == "Yes" ~ 1
  )) %>%
  dplyr::mutate(seminal_vesicle_invasion = case_when(
    uncurated$`svi:ch1` == "Negative" ~ 0,
    uncurated$`svi:ch1` == "Positive" ~ 1
  )) %>%
  dplyr::mutate(genome_altered = uncurated$"%cna (percent of  genome that is copy number altered):ch1") %>%
  dplyr::mutate(therapy_radiation_initial = dplyr::case_when(
    is.na(uncurated$"adjradtx:ch1") ~ 0,
    TRUE ~ 1
  )) %>%
  dplyr::mutate(therapy_hormonal_initial = dplyr::case_when(
    is.na(uncurated$"adjhormtx:ch1") ~ 0,
    TRUE ~ 1
  )) %>%
  dplyr::mutate(other_treatment = dplyr::case_when(
    is.na(uncurated$"adjchemotx:ch1") ~ "",
    uncurated$"adjchemotx:ch1" == "CLIN_TRIAL" ~ "Clinical trial chemo",
    TRUE ~ ""
  )) %>%
  dplyr::mutate(tumor_margins_positive = dplyr::case_when(
    uncurated$"sms:ch1" == "Positive" ~ 1,
    TRUE ~ 0
  ))

rownames(curated) <- sub(".*\\s", "", uncurated$title)
clinical_hieronymus <- curated

save(clinical_hieronymus, file = "data-raw/clinical_hieronymus.RData")

###
#
# ICGC clinical data curation scripts
# Downloaded from ICGC's data releases for clinical metadata
#
###

###
## CA (Canadian) ICGC dataset
###

# Generation script from generate.R, same function as is used for downloading and transforming the omics
uncurated <- curatedPCaData:::generate_icgc("PRAD_CA", "clinical")

# Format empty data frames according to the prad template
curated <- initial_curated_df(
  df_rownames = uncurated$icgc_sample_id,
  template_name = "data-raw/template_prad.csv"
)

# Curate available fields
curated <- curated %>%
  dplyr::mutate(study_name = "ICGC_CA") %>%
  dplyr::mutate(sample_name = uncurated$icgc_sample_id) %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated$specimen_type == "Normal - blood derived" ~ "normal",
    uncurated$specimen_type == "Primary tumour - solid tissue" ~ "primary"
  )) %>%
  dplyr::mutate(patient_id = uncurated$icgc_donor_id) %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    uncurated$donor_vital_status == "alive" ~ 0,
    uncurated$donor_vital_status == "deceased" ~ 1,
    uncurated$donor_vital_status == "" ~ NA_real_,
    is.na(uncurated$donor_vital_status) ~ NA_real_
  )) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$donor_age_at_diagnosis) %>%
  dplyr::mutate(days_to_overall_survival = (uncurated$donor_age_at_last_followup - uncurated$donor_age_at_diagnosis) * 365) %>%
  dplyr::mutate(gleason_major = as.integer(stringr::str_sub(uncurated$tumour_grade, 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.integer(stringr::str_sub(uncurated$tumour_grade, 3, 3))) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    stringr::str_sub(uncurated$tumour_grade, 1, 3) == "3+3" ~ "<=6",
    stringr::str_sub(uncurated$tumour_grade, 1, 3) == "3+4" ~ "3+4",
    stringr::str_sub(uncurated$tumour_grade, 1, 3) == "4+3" ~ "4+3",
    stringr::str_sub(uncurated$tumour_grade, 1, 3) %in% c("4+4", "4+5") ~ ">=8"
  )) %>%
  dplyr::mutate(gleason_grade = gleason_minor + gleason_major) %>%
  dplyr::mutate(T_clinical = readr::parse_number(uncurated$tumour_stage)) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(uncurated$tumour_stage, "[a-c]+"))

clinical_icgcca <- curated

save(clinical_icgcca, file = "data-raw/clinical_icgcca.RData")

###
## FR (French) ICGC dataset
###

# Generation script from generate.R, same function as is used for downloading and transforming the omics
uncurated <- curatedPCaData:::generate_icgc("PRAD_FR", "clinical")

# Format empty data frames according to the prad template
curated <- initial_curated_df(
  df_rownames = rownames(uncurated),
  template_name = "data-raw/template_prad.csv"
)

# Mimic previous curation piping
curated <- curated %>%
  dplyr::mutate(study_name = "ICGC_FR") %>%
  dplyr::mutate(sample_name = uncurated$icgc_sample_id) %>%
  dplyr::mutate(patient_id = uncurated$icgc_donor_id) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$donor_age_at_diagnosis) %>%
  dplyr::mutate(days_to_overall_survival = uncurated$donor_survival_time) %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    uncurated$donor_vital_status == "alive" ~ 0,
    uncurated$donor_vital_status == "deceased" ~ 1,
    is.na(uncurated$donor_vital_status) ~ NA_real_
  ))
# Source claims to use GLEASON_SCORE and PT_STAGE fields, yet the values in corresponding columns are NA?
# TODO: Verify if these are at some other table

clinical_icgcfr <- curated

save(clinical_icgcfr, file = "data-raw/clinical_icgcfr.RData")

###
## UK (United Kingdom) ICGC dataset
###

# Generation script from generate.R, same function as is used for downloading and transforming the omics
uncurated <- curatedPCaData:::generate_icgc("PRAD_UK", "clinical")

# Format empty data frames according to the prad template
curated <- initial_curated_df(
  df_rownames = rownames(uncurated),
  template_name = "data-raw/template_prad.csv"
)

# Mimic previous curation piping
curated <- curated %>%
  dplyr::mutate(study_name = "ICGC_UK") %>%
  dplyr::mutate(sample_name = uncurated$icgc_sample_id) %>%
  dplyr::mutate(patient_id = uncurated$icgc_donor_id) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$donor_age_at_diagnosis) %>%
  dplyr::mutate(days_to_overall_survival = uncurated$donor_survival_time) %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    uncurated$donor_vital_status == "alive" ~ 0,
    uncurated$donor_vital_status == "deceased" ~ 1,
    uncurated$donor_vital_status == "" ~ NA_real_,
    is.na(uncurated$donor_vital_status) ~ NA_real_
  )) %>%
  dplyr::mutate(gleason_major = as.integer(stringr::str_sub(uncurated$tumour_grade, 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.integer(stringr::str_sub(uncurated$tumour_grade, 3, 3))) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    stringr::str_sub(uncurated$tumour_grade, 1, 3) == "3+3" ~ "<=6",
    stringr::str_sub(uncurated$tumour_grade, 1, 3) == "3+4" ~ "3+4",
    stringr::str_sub(uncurated$tumour_grade, 1, 3) == "4+3" ~ "4+3",
    stringr::str_sub(uncurated$tumour_grade, 1, 3) %in% c("4+4", "4+5") ~ ">=8"
  )) %>%
  dplyr::mutate(gleason_grade = gleason_minor + gleason_major) %>%
  dplyr::mutate(metastasis_occurrence_status = dplyr::case_when(
    stringr::str_sub(uncurated$tumour_stage, 6, 7) == "M1" ~ 1,
    stringr::str_sub(uncurated$tumour_stage, 6, 7) == "M0" ~ 0,
    stringr::str_sub(uncurated$tumour_stage, 6, 7) %in% c("Mx", "") ~ NA_real_
  )) %>%
  dplyr::mutate(M_stage = stringr::str_sub(uncurated$tumour_stage, 6, 7))
# clinical -> uncurated$donor_tumour_stage_at_diagnosis
# pathology -> uncurated$tumour_stage
## contains a lot of information for the TNM
#> table(uncurated$donor_tumour_stage_at_diagnosis)
#
#         pT2c pN0  pT3 pN0 pT3b pN1  pT4 pN1  T1bN0Mx  T1cN0M0  T1cN0Mx
#       5        5       10        6        4        2        4       18
# T1cNxM1  T1cNxMx    T2 Nx  T2aN0Mx  T2aNxMx  T2b pN1  T2bN0M0  T2bN0Mx
#       2      164        3       45       20        6        2       14
# T2bNxMx  T2cN0M0  T2cN0M1  T2cN0Mx  T2cN1Mx  T2cNxM0  T2cNxMx   T2N0Mx
#       8        8        2       50        2        2        8        2
#  T2NxMx    T3 Nx  T3aN0M0  T3aN0Mx  T3aNxMx  T3bN0Mx  T3bNxMx   T3N0Mx
#       6        5        2       20       23        2        6        2
#  T3NxM0    T4 Nx   TxNxMx
#       2       17       32
#> table(uncurated$tumour_stage)
#
#        T2aNxMx T2cN0M0 T2cN0M1 T2cN0Mx T2cNxMx T3aN0M0 T3aN0Mx T3aN1Mx T3aNxMx
#    273       6       8       2      20      44       6      65       2      37
# T3bN0Mx T3bN1M0 T3bN1Mx T3bNxMx T4xNxMx TxxNxMx
#   12       2      12       4       2      14
#

clinical_icgcuk <- curated

save(clinical_icgcuk, file = "data-raw/clinical_icgcuk.RData")

###
#
# IGC - GSE2109
# Source: GEO (subset just to the prostate cancer samples)
#
###
gse <- GEOquery::getGEO("GSE2109", GSEMatrix = TRUE)

uncurated <- Biobase::pData(gse[[1]])
uncurated <- uncurated[grep("Prostate", uncurated$title), ]

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)
# uncurated <- uncurated$title

uncurated_grep <- data.frame(
  title = uncurated$title,
  geo = uncurated$geo_accession,
  study = "IGC",
  # Grep through the data row-wise; data is broken into wrong columns, but value prefixes are correct
  ethnicity = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Ethnic Background: ", "", grep("Ethnic Background: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  tobacco_use = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Tobacco Use : ", "", grep("Tobacco Use : ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  # A range of values; e.g. 50-60, 60-70, ...
  age = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Patient Age: ", "", grep("Patient Age: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  psa = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("PSA: ", "", grep("PSA: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  # Redundant, all 'Prostate' samples in this case, no metastases
  # unlist(apply(uncurated, MARGIN=1, FUN=function(x){
  # 	tmp <- gsub("Primary Site: ", "", grep("Primary Site: ", x, value=TRUE)); ifelse(length(tmp)>0, as.character(tmp), NA)
  # }))
  histology = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Histology: ", "", grep("Histology: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  path_T = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Pathological T: ", "", grep("Pathological T: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  path_M = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Pathological M: ", "", grep("Pathological M: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.numeric(tmp), NA)
  })),
  path_N = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Pathological N: ", "", grep("Pathological N: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.numeric(tmp), NA)
  })),
  path_stage = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Pathological Stage: ", "", grep("Pathological Stage: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  path_gleason = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Pathological Gleason Score: ", "", grep("Pathological Gleason Score: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  clin_T = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Clinical T: ", "", grep("Clinical T: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  clin_N = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Clinical N: ", "", grep("Clinical N: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.numeric(tmp), NA)
  })),
  clin_M = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Clinical M: ", "", grep("Clinical M: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.numeric(tmp), NA)
  })),
  clin_stage = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Clinical Stage: ", "", grep("Clinical Stage: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  })),
  clin_gleason = unlist(apply(uncurated, MARGIN = 1, FUN = function(x) {
    tmp <- gsub("Clinical Gleason Score: ", "", grep("Clinical Gleason Score: ", x, value = TRUE))
    ifelse(length(tmp) > 0, as.character(tmp), NA)
  }))
)

# Re-generate IGC based on the grep-failsafe'd data structure
curated <- curated %>%
  dplyr::mutate(study_name = "IGC") %>%
  dplyr::mutate(sample_name = rownames(uncurated_grep)) %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(patient_id = rownames(uncurated_grep)) %>%
  dplyr::mutate(alt_sample_name = uncurated_grep$title) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(race = dplyr::case_when(
    uncurated_grep$ethnicity == "Caucasian" ~ "caucasian",
    uncurated_grep$ethnicity == "African-American" ~ "african-american",
    uncurated_grep$ethnicity == "American Indian" ~ "other",
    uncurated_grep$ethnicity == "Asian" ~ "asian",
    TRUE ~ NA_character_
  )) %>%
  dplyr::mutate(smoking_status = as.numeric(uncurated_grep$tobacco_use == "Yes")) %>%
  # Take the middle value of age range as the representative integer
  dplyr::mutate(age_at_initial_diagnosis = round(unlist(lapply(uncurated_grep$age, FUN = function(x) {
    mean(as.numeric(strsplit(x, "-")[[1]]))
  })), 0)) %>%
  # Save histological subgroup
  dplyr::mutate(other_sample = paste0("histology=", gsub("Adenocarcinoma, ", "", uncurated_grep$histology))) %>%
  # Clinical info
  dplyr::mutate(T_clinical = as.numeric(substr(uncurated_grep$clin_T, 1, 1))) %>%
  dplyr::mutate(T_substage_clinical = as.character(substr(uncurated_grep$clin_T, 2, 2))) %>%
  # Pathological info
  dplyr::mutate(T_pathological = as.numeric(substr(uncurated_grep$path_T, 1, 1))) %>%
  dplyr::mutate(T_substage_pathological = as.character(substr(uncurated_grep$path_T, 2, 2))) %>%
  # Only clinical or pathological grade group was available; pick the one that was available for a patient
  dplyr::mutate(grade_group = ifelse(is.na(uncurated_grep$path_gleason), uncurated_grep$clin_gleason, uncurated_grep$path_gleason)) %>%
  # Change representation
  dplyr::mutate(grade_group = dplyr::case_when(
    grade_group == "5-6" ~ "<=6",
    grade_group == "7" ~ "7", # Unfortunately 4+3 and 3+4 collapsed to a single value
    grade_group == "8-10" ~ ">=8",
    TRUE ~ NA_character_
  )) %>%
  # Metastases from either pathology or clinical field
  dplyr::mutate(M_stage = ifelse(is.na(uncurated_grep$path_M), uncurated_grep$clin_M, uncurated_grep$path_M)) %>%
  # Lymph nodes from either pathology or clinical field
  dplyr::mutate(N_stage = ifelse(is.na(uncurated_grep$path_N), uncurated_grep$clin_N, uncurated_grep$path_N)) %>%
  # PSA was only reported as within 'Normal' range or 'Elevated'
  dplyr::mutate(psa_category = uncurated_grep$psa)

clinical_igc <- curated

save(clinical_igc, file = "data-raw/clinical_igc.RData")

###
#
# Kim et al.
# curated from GEO's clinical metadata (GSM)
#
###

gse <- GEOquery::getGEO("GSE119616", GSEMatrix = TRUE)

uncurated <- Biobase::pData(gse[[1]])

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Kim et al.") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(patient_id = stringr::str_remove(
    uncurated$title,
    "prostate_cancer_biopsy_sample_pid_"
  )) %>%
  dplyr::mutate(tissue_source = gsub("prostate cancer biopsy", "biopsy", stringr::str_remove(uncurated$characteristics_ch1.10, "tissue:"))) %>%
  dplyr::mutate(age_at_initial_diagnosis = floor(as.numeric(uncurated$`age:ch1`))) %>%
  dplyr::mutate(psa = as.numeric(stringr::str_remove(uncurated$characteristics_ch1.6, "psa:"))) %>%
  dplyr::mutate(gleason_major = as.numeric(stringr::str_remove(uncurated$characteristics_ch1.8, "primary gleason grade:"))) %>%
  dplyr::mutate(gleason_minor = as.numeric(stringr::str_remove(uncurated$characteristics_ch1.9, "secondary gleason grade:"))) %>%
  dplyr::mutate(gleason_grade = as.numeric(gleason_major + gleason_minor)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade %in% 5:6 ~ "<=6",
    gleason_major == "3" & gleason_minor == "4" ~ "3+4",
    gleason_major == "4" & gleason_minor == "3" ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8"
  )) %>%
  dplyr::mutate(T_clinical = readr::parse_number(uncurated$`tumor stage:ch1`)) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(uncurated$`tumor stage:ch1`, "[a-c]+")) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(frozen_ffpe = "ffpe") %>%
  dplyr::mutate(batch = uncurated$"institution:ch1") %>%
  dplyr::mutate(tissue_source = "biopsy")

# Non-standard fields should be added as 'other features' or similar fields depending on their type
curated$other_sample <- apply(cbind(
  paste0("nccn=", uncurated$"nccn:ch1"),
  paste0("percent_positive_cores=", round(as.numeric(uncurated$"percent positive cores:ch1"), 4))
), MARGIN = 1, FUN = function(x) {
  paste(x, collapse = "|")
})

clinical_kim <- curated

save(clinical_kim, file = "data-raw/clinical_kim.RData")

###
#
# Kunderfranco et al.
# curated from GEO's clinical metadata (GSM)
#
###

gset <- GEOquery::getGEO("GSE14206", GSEMatrix = TRUE, getGPL = TRUE)

# clinical
uncurated <- Biobase::pData(gset[[1]])

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Kunderfranco et al.") %>%
  dplyr::mutate(sample_name = uncurated$geo_accession) %>%
  dplyr::mutate(patient_id = stringr::str_sub(uncurated$description.1, 22, 30)) %>%
  dplyr::mutate(alt_sample_name = uncurated$title) %>%
  dplyr::mutate(gleason_major = as.numeric(stringr::str_sub(uncurated$"gleason grade:ch1", 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.numeric(stringr::str_sub(uncurated$"gleason grade:ch1", 3, 3))) %>%
  dplyr::mutate(gleason_grade = gleason_major + gleason_minor) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade %in% 4:6 ~ "<=6",
    gleason_major == 3 & gleason_minor == 4 ~ "3+4",
    gleason_major == 4 & gleason_minor == 3 ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8",
    TRUE ~ "NA"
  )) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$"age:ch1") %>%
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    ## TDL: Normal samples were not BPH!
    uncurated$source_name_ch1 == "prostate cancer" ~ "primary",
    # uncurated$source_name_ch1 == "normal prostate" ~ 'BPH')) %>%
    uncurated$source_name_ch1 == "normal prostate" ~ "normal"
  )) %>%
  dplyr::mutate(frozen_ffpe = "FFPE") %>%
  dplyr::mutate(source_of_gleason = dplyr::case_when(
    uncurated$source_name_ch1 == "prostate cancer" ~ "prostatectomy",
    uncurated$source_name_ch1 == "normal prostate" ~ "biopsy"
  )) %>%
  dplyr::mutate(microdissected = 0) %>%
  dplyr::mutate(tissue_source = dplyr::case_when(
    uncurated$source_name_ch1 == "prostate cancer" ~ "prostatectomy",
    uncurated$source_name_ch1 == "normal prostate" ~ "biopsy"
  )) %>%
  dplyr::mutate(other_sample = paste0("ets_group=", uncurated$"ets group:ch1")) %>%
  dplyr::mutate(T_clinical = as.numeric(substr(uncurated$"stage:ch1", 2, 2))) %>%
  dplyr::mutate(T_substage_clinical = base::tolower(substr(uncurated$"stage:ch1", 3, 3)))

clinical_kunderfranco <- curated

save(clinical_kunderfranco, file = "./data-raw/clinical_kunderfranco.RData")


###
#
# Ren et al. eururol 2017
# curated from cgdsr
#
###

mycgds <- cgdsr::CGDS("http://www.cbioportal.org/")
uncurated <- cgdsr::getClinicalData(mycgds, caseList = "prad_eururol_2017_sequenced")
# TDL: Some fields were missing from cBioDataPack but were available with cgdsr
# mae <- cBioPortalData::cBioDataPack("prad_eururol_2017",ask = FALSE)
# uncurated2 <- MultiAssayExperiment::colData(mae)
# uncurated2 <- as.data.frame(uncurated)

# Create the curated object templates (cgdsr and cBioPortalData separately, merge later)
curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Ren et al.") %>%
  dplyr::mutate(patient_id = rownames(uncurated)) %>%
  dplyr::mutate(age_at_initial_diagnosis = as.numeric(uncurated$AGE)) %>%
  dplyr::mutate(sample_name = rownames(uncurated)) %>%
  # From the publication: "The study sequenced whole-genome and transcriptome of tumor-benign paired tissues from 65 treatment-naive Chinese PCa patients"
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(race = "asian") %>%
  dplyr::mutate(patient_id = row.names(uncurated)) %>%
  dplyr::mutate(psa = as.numeric(uncurated$PSA)) %>%
  dplyr::mutate(gleason_grade = as.numeric(stringr::str_sub(uncurated$GLEASON_SCORE, 1, 1)) + as.numeric(stringr::str_sub(uncurated$GLEASON_SCORE, 3, 3))) %>%
  dplyr::mutate(gleason_major = as.numeric(stringr::str_sub(uncurated$GLEASON_SCORE, 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.numeric(stringr::str_sub(uncurated$GLEASON_SCORE, 3, 3))) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade %in% 2:6 ~ "<=6",
    stringr::str_sub(uncurated$GLEASON_SCORE, 1, 3) == "3+4" ~ "3+4",
    stringr::str_sub(uncurated$GLEASON_SCORE, 1, 3) == "4+3" ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8"
  )) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$AGE) %>%
  dplyr::mutate(T_clinical = readr::parse_number(uncurated$TNMSTAGE)) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(uncurated$TNMSTAGE, "[a-c]+")) %>%
  # dplyr::mutate(FPSA_PSA = uncurated$FPSA_PSA) %>%
  # Tumor purity was estimated using ASCAT (Allele-Specific Copy number Analysis of Tumours)
  dplyr::mutate(tumor_purity_ascat = uncurated$TUMOR_PURITY) %>%
  # dplyr::mutate(BLADDER_NECK_INVASION = uncurated$BLADDER_NECK_INVASION) %>%
  dplyr::mutate(seminal_vesicle_invasion = dplyr::case_when(
    uncurated$SEMINAL_VESICLE_INVASION == "No" ~ 0,
    uncurated$SEMINAL_VESICLE_INVASION == "Yes" ~ 1,
    TRUE ~ NA_real_
  )) %>%
  dplyr::mutate(genome_altered = uncurated$FRACTION_GENOME_ALTERED) %>%
  dplyr::mutate(N_stage = dplyr::case_when(
    uncurated$LYMPH_NODE_METASTASIS == "No" ~ 0,
    uncurated$LYMPH_NODE_METASTASIS == "Yes" ~ 1,
    uncurated$LYMPH_NODE_METASTASIS == "" ~ NA_real_
  )) # %>%
# dplyr::mutate(MUTATION_COUNT=uncurated$MUTATION_COUNT)

# Non-standard fields should be added as 'other features' or similar fields depending on their type
curated$other_sample <- apply(cbind(
  paste0("FPSA_PSA=", uncurated$FPSA_PSA),
  paste0("BLADDER_NECK_INVASION=", uncurated$BLADDER_NECK_INVASION),
  paste0("MUTATION_COUNT=", uncurated$MUTATION_COUNT),
  paste0("TMB_NONSYNONYMOUS=", round(uncurated$TMB_NONSYNONYMOUS, 4))
), MARGIN = 1, FUN = function(x) {
  paste(x, collapse = "|")
})

rownames(curated) <- curated$patient_id

clinical_ren <- curated

save(clinical_ren, file = "data-raw/clinical_ren.RData")

###
#
# Sun et al.
# from GEOquery's clinical data fields (GSM)
#
###

gse <- GEOquery::getGEO("GSE25136", GSEMatrix = TRUE)

uncurated <- Biobase::pData(gse[[1]])

curated <- initial_curated_df(
  df_rownames = rownames(uncurated),
  template_name = "data-raw/template_prad.csv"
)

curated <- curated %>%
  dplyr::mutate(study_name = "Sun, et al.") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(patient_id = row.names(uncurated)) %>%
  dplyr::mutate(alt_sample_name = stringr::str_remove(
    uncurated$title,
    "Prostate cancer primary tumor "
  )) %>%
  dplyr::mutate(disease_specific_recurrence_status = stringr::str_remove(
    uncurated$characteristics_ch1.1,
    "recurrence status: "
  )) %>%
  dplyr::mutate(disease_specific_recurrence_status = dplyr::case_when(
    disease_specific_recurrence_status == "Non-Recurrent" ~ 0,
    disease_specific_recurrence_status == "Recurrent" ~ 1,
    TRUE ~ NA_real_
  )) %>%
  dplyr::mutate(sample_type = stringr::str_remove(
    uncurated$`tissue:ch1`,
    "Prostate cancer "
  )) %>%
  dplyr::mutate(sample_type = stringr::str_remove(sample_type, " tumor"))

clinical_sun <- curated

save(clinical_sun, file = "data-raw/clinical_sun.RData")

###
#
# Taylor et al.
# Clinical metadata combined from GEO's clinical metadata (GSM) and metadata from cBioPortalData
#
###

gse_all <- GEOquery::getGEO("GSE21032", GSEMatrix = TRUE)

gse <- rbind(
  # Transcript
  Biobase::pData(gse_all[[1]])[, c("sample id:ch1", "tissue:ch1", "tumor type:ch1", "biopsy_gleason_grade:ch1", "clint_stage:ch1", "disease status:ch1", "pathological_stage:ch1")],
  # aCGH
  Biobase::pData(gse_all[[2]])[, c("sample id:ch1", "tissue:ch1", "tumor type:ch1", "biopsy_gleason_grade:ch1", "clint_stage:ch1", "disease status:ch1", "pathological_stage:ch1")],
  # Exon
  Biobase::pData(gse_all[[3]])[, c("sample id:ch1", "tissue:ch1", "tumor type:ch1", "biopsy_gleason_grade:ch1", "clint_stage:ch1", "disease status:ch1", "pathological_stage:ch1")],
  # miRNA
  Biobase::pData(gse_all[[4]])[, c("sample id:ch1", "tissue:ch1", "tumor type:ch1", "biopsy_gleason_grade:ch1", "clint_stage:ch1", "disease status:ch1", "pathological_stage:ch1")]
)

mae <- cBioPortalData::cBioDataPack("prad_mskcc", ask = FALSE)
uncurated <- MultiAssayExperiment::colData(mae)
uncurated_cbio <- as.data.frame(uncurated)

curated <- initial_curated_df(
  df_rownames = gse$"sample id:ch1",
  template_name = "data-raw/template_prad.csv"
)

curated <- curated %>%
  dplyr::mutate(study_name = "Taylor, et al.") %>%
  # Save GEO accession codes
  dplyr::mutate(alt_sample_name = rownames(gse)) %>%
  # dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(patient_id = gse$"sample id:ch1") %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(sample_type = tolower(gse$"tumor type:ch1")) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    sample_type == "primary tumor" ~ "primary",
    sample_type == "cell line" ~ "cell.line",
    sample_type == "xenograft" ~ "xenograft",
    sample_type == "metastatsis" ~ "metastatic",
    is.na(sample_type) ~ "normal",
    TRUE ~ sample_type
  )) %>%
  # Clinical substaging
  dplyr::mutate(gleason_grade = as.numeric(gse$"biopsy_gleason_grade:ch1")) %>%
  dplyr::mutate(T_clinical = stringr::str_sub(gse$"clint_stage:ch1", 2, 2)) %>%
  dplyr::mutate(T_clinical = as.numeric(T_clinical)) %>%
  dplyr::mutate(T_substage_clinical = tolower(stringr::str_sub(gse$"clint_stage:ch1", 3, 3))) %>%
  dplyr::mutate(T_substage_clinical = dplyr::case_when(
    T_substage_clinical == "" ~ NA_character_,
    TRUE ~ T_substage_clinical
  )) %>%
  # Pathology substaging
  dplyr::mutate(T_pathological = as.numeric(stringr::str_sub(gse$"pathological_stage:ch1", 2, 2))) %>%
  dplyr::mutate(T_substage_pathological = tolower(stringr::str_sub(gse$"pathological_stage:ch1", 3, 3))) %>%
  dplyr::mutate(T_substage_pathological = dplyr::case_when(
    T_substage_pathological == "" ~ NA_character_,
    TRUE ~ T_substage_pathological
  ))

# Order additional clinical information from cBio according to the order in current df
uncurated_cbio <- uncurated_cbio[match(curated$patient_id, row.names(uncurated_cbio)), ]

# Additional clinical parameters as recored in cBioPortal
curated <- curated %>%
  # Gleason grades reported in GEO seem to differ from cBioPortal even for same IDs; using the ones provided by cBio:
  dplyr::mutate(gleason_grade = as.integer(uncurated_cbio$GLEASON_SCORE)) %>%
  dplyr::mutate(gleason_major = as.integer(uncurated_cbio$GLEASON_SCORE_1)) %>%
  dplyr::mutate(gleason_minor = as.integer(uncurated_cbio$GLEASON_SCORE_2)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade == 6 ~ "<=6",
    gleason_grade %in% 8:10 ~ ">=8",
    gleason_major == 3 & gleason_minor == 4 ~ "3+4",
    gleason_major == 4 & gleason_minor == 3 ~ "4+3",
  )) %>%
  dplyr::mutate(ERG_fusion_GEX = uncurated_cbio$ERG_FUSION_GEX) %>%
  dplyr::mutate(ERG_fusion_GEX = dplyr::case_when(
    ERG_fusion_GEX == "Negative" ~ 0,
    ERG_fusion_GEX == "Positive" ~ 1,
    TRUE ~ NA_real_
  )) %>%
  dplyr::mutate(ERG_fusion_CNA = uncurated_cbio$ERG_FUSION_ACGH) %>%
  dplyr::mutate(ERG_fusion_CNA = dplyr::case_when(
    ERG_fusion_CNA == "Positive" ~ 1,
    ERG_fusion_CNA %in% c("Negative", "Flat") ~ 0,
    TRUE ~ NA_real_
  )) %>%
  dplyr::mutate(disease_specific_recurrence_status = uncurated_cbio$DFS_STATUS) %>%
  dplyr::mutate(disease_specific_recurrence_status = dplyr::case_when(
    disease_specific_recurrence_status == "1:Recurred" ~ 1,
    disease_specific_recurrence_status == "0:DiseaseFree" ~ 0,
    TRUE ~ NA_real_
  )) %>%
  dplyr::mutate(days_to_disease_specific_recurrence = round(uncurated_cbio$DFS_MONTHS * 30.5, 0)) %>%
  dplyr::mutate(genome_altered = uncurated_cbio$FRACTION_GENOME_ALTERED)

# Leave out cell.line and xenograft samples
curated <- curated[which(curated$sample_type %in% c("primary", "metastatic", "normal")), ]
curated <- curated[grep("PCA|PAN", curated$sample_name), ]
curated <- curated[order(curated$sample_name), ]
# Only include unique entries
curated <- curated[which(!duplicated(curated$patient_id)), ]

rownames(curated) <- curated$sample_name
clinical_taylor <- curated

save(clinical_taylor, file = "data-raw/clinical_taylor.RData")

###
#
# TCGA dataset
# Clinical metadata combined from multiple sources/datasets:
# cgdsr (cBioPortal) patient and sample level and Xena Hub's clinical metadata
#
###

mycgds <- cgdsr::CGDS("http://www.cbioportal.org/")
uncurated <- cgdsr::getClinicalData(mycgds, caseList = "prad_tcga_sequenced")

curated <- initial_curated_internal(df_rownames = rownames(uncurated))

curated <- curated %>%
  # Not just TCGA provisional with TCGA firehose brought in for GEX
  dplyr::mutate(study_name = "TCGA") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(frozen_ffpe = uncurated$IS_FFPE) %>%
  dplyr::mutate(frozen_ffpe = dplyr::case_when(
    frozen_ffpe %in% c("NO", "[Not Available]") ~ "NA",
    frozen_ffpe == "YES" ~ "ffpe",
    TRUE ~ frozen_ffpe
  )) %>%
  # Somatic differences had paired samples available, GEX was unpaired
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated$SAMPLE_TYPE == "Metastasis" ~ "metastatic",
    uncurated$SAMPLE_TYPE == "Primary" ~ "primary"
  )) %>%
  dplyr::mutate(patient_id = stringr::str_sub(sample_name, 1, 12)) %>%
  dplyr::mutate(alt_sample_name = uncurated$OTHER_SAMPLE_ID) %>%
  dplyr::mutate(gleason_grade = uncurated$GLEASON_SCORE) %>%
  dplyr::mutate(gleason_major = uncurated$GLEASON_PATTERN_PRIMARY) %>%
  dplyr::mutate(gleason_minor = uncurated$GLEASON_PATTERN_SECONDARY) %>%
  # dplyr::mutate(source_of_gleason = "biopsy") %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade == 6 ~ "<=6",
    gleason_major == 3 & gleason_minor == 4 ~ "3+4",
    gleason_major == 4 & gleason_minor == 3 ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8",
    TRUE ~ "NA"
  )) %>%
  dplyr::mutate(zone_of_origin = uncurated$PRIMARY_SITE) %>%
  dplyr::mutate(zone_of_origin = dplyr::case_when(
    zone_of_origin == "Peripheral Zone" ~ "peripheral",
    zone_of_origin == "Overlapping / Multiple Zones" ~ "mixed",
    zone_of_origin == "Central Zone" ~ "central",
    zone_of_origin == "Transition Zone" ~ "transitional",
    zone_of_origin == "[Not Available]" ~ "NA",
    TRUE ~ "NA"
  )) %>%
  dplyr::mutate(year_diagnosis = uncurated$INITIAL_PATHOLOGIC_DX_YEAR) %>%
  dplyr::mutate(overall_survival_status = uncurated$OS_STATUS) %>%
  dplyr::mutate(overall_survival_status = dplyr::case_when(
    is.na(overall_survival_status) ~ NA_real_,
    overall_survival_status == "1:DECEASED" ~ 1,
    overall_survival_status == "0:LIVING" ~ 0
  )) %>%
  dplyr::mutate(days_to_overall_survival = as.numeric(uncurated$OS_MONTHS) * 30.5) %>%
  dplyr::mutate(disease_specific_recurrence_status = uncurated$DFS_STATUS) %>%
  dplyr::mutate(disease_specific_recurrence_status = dplyr::case_when(
    disease_specific_recurrence_status == "" ~ NA_real_,
    disease_specific_recurrence_status == "1:Recurred/Progressed" ~ 1,
    disease_specific_recurrence_status == "0:DiseaseFree" ~ 0
  )) %>%
  dplyr::mutate(days_to_disease_specific_recurrence = as.numeric(uncurated$DFS_MONTHS) * 30.5) %>%
  dplyr::mutate(psa = uncurated$PSA_MOST_RECENT_RESULTS) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated$AGE) %>%
  dplyr::mutate(M_stage = uncurated$CLIN_M_STAGE) %>%
  dplyr::mutate(M_stage = dplyr::case_when(
    M_stage == "M0" ~ 0,
    M_stage %in% c("M1a", "M1b", "M1c") ~ 1,
    TRUE ~ NA_real_
  )) %>%
  # stringr:: commands return true NA not character NA
  dplyr::mutate(M_substage = stringr::str_sub(uncurated$CLIN_M_STAGE, 3, 3)) %>%
  # single instance '[Unknown]' will throw a warning for below
  dplyr::mutate(T_clinical = readr::parse_number(uncurated$CLIN_T_STAGE)) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(uncurated$CLIN_T_STAGE, "[a-c]+")) %>%
  dplyr::mutate(T_pathological = readr::parse_number(uncurated$PATH_T_STAGE)) %>%
  dplyr::mutate(T_substage_pathological = stringr::str_extract(uncurated$PATH_T_STAGE, "[a-c]+")) %>%
  dplyr::mutate(race = uncurated$RACE) %>%
  dplyr::mutate(race = dplyr::case_when(
    race == "WHITE" ~ "caucasian",
    race == "ASIAN" ~ "asian",
    race == "BLACK OR AFRICAN AMERICAN" ~ "african_american",
    TRUE ~ "NA"
  )) %>%
  dplyr::mutate(therapy_radiation_initial = uncurated$RADIATION_TREATMENT_ADJUVANT) %>%
  # Radiation treatment given at initial treatment
  dplyr::mutate(therapy_radiation_initial = dplyr::case_when(
    therapy_radiation_initial == "YES" ~ 1,
    therapy_radiation_initial == "NO" ~ 0,
    therapy_radiation_initial == "" ~ NA_real_
  )) %>%
  dplyr::mutate(other_treatment = uncurated$TARGETED_MOLECULAR_THERAPY) %>%
  # Add additional treatments
  dplyr::mutate(other_treatment = dplyr::case_when(
    other_treatment == "YES" ~ 1,
    other_treatment == "NO" ~ 0,
    other_treatment == "" ~ NA_real_
  )) %>%
  # Fraction of genome altered
  dplyr::mutate(genome_altered = uncurated$FRACTION_GENOME_ALTERED) %>%
  # Nx, N0 or N1 if findings in lymph nodes
  dplyr::mutate(N_stage = uncurated$PATH_N_STAGE) %>%
  dplyr::mutate(N_stage = dplyr::case_when(
    N_stage == "N0" ~ 0,
    N_stage == "N1" ~ 1,
    N_stage == "" ~ NA_real_
  ))


# Append additional information from Firehose Legacy; loop over all 'omics profiles
uncurated2 <- do.call("rbind", lapply(cgdsr::getGeneticProfiles(mycgds, "prad_tcga_pub")[, 1], FUN = function(x) {
  cgdsr::getClinicalData(mycgds, caseList = x)
}))
# Differences in reported fields (already based on GEX)
# > length(uncurated)
# [1] 80
# > length(uncurated2)
# [1] 91
#
curated2 <- initial_curated_internal(df_rownames = rownames(uncurated2))

# Match rownames between subsets
uncurated2 <- uncurated2[match(rownames(uncurated), rownames(uncurated2)), ]

curated2 <- uncurated2 %>%
  # ERG fusions were determined using gene expression
  dplyr::mutate(ERG_fusion_GEX = dplyr::case_when(
    ERG_STATUS == "fusion" ~ 1,
    ERG_STATUS == "none" ~ 0,
    is.na(ERG_STATUS) ~ NA_real_
  )) %>%
  # Purity estimates from TCGA Firehose legacy
  dplyr::mutate(tumor_purity_absolute = uncurated2$ABSOLUTE_EXTRACT_PURITY) %>%
  dplyr::mutate(tumor_purity_demixt = uncurated2$DEMIX_PURITY) %>%
  dplyr::mutate(tumor_purity_pathology = uncurated2$TUMOR_CELLULARITY_PATHOLOGY_REVIEW) %>%
  # Precomputed AR-scores
  dplyr::mutate(AR_activity = uncurated2$AR_SCORE) %>%
  # Ethnicity
  dplyr::mutate(race = dplyr::case_when(
    RACE == "ASIAN" ~ "asian",
    RACE == "BLACK OR AFRICAN AMERICAN" ~ "african_american",
    RACE == "WHITE" ~ "caucasian",
    is.na(RACE) ~ NA_character_
  ))

# Copy additional curated field from Firehose legacy
curated[, "ERG_fusion_GEX"] <- curated2[, "ERG_fusion_GEX"]
curated[, "tumor_purity_absolute"] <- curated2[, "tumor_purity_absolute"]
curated[, "tumor_purity_demixt"] <- curated2[, "tumor_purity_demixt"]
curated[, "tumor_purity_pathology"] <- curated2[, "tumor_purity_pathology"]
curated[, "AR_activity"] <- curated2[, "AR_activity"]
curated[, "race"] <- curated2[, "race"]

# Subset to analytical subset of 333 samples prior to removing low quality samples
curated <- curated[which(curated$sample_name %in% intersect(rownames(uncurated), rownames(uncurated2))), ]
rownames(curated) <- curated$sample_name

# TCGA - Xenabrowser
# Re-run of clinical fields

uncurated4 <- curatedPCaData:::generate_xenabrowser(
  id = "TCGA-PRAD",
  type = "clinical",
  truncate = 0
)

curated4 <- initial_curated_internal(df_rownames = gsub("-", ".", rownames(uncurated4)))

curated4 <- curated4 %>%
  dplyr::mutate(study_name = "TCGA") %>%
  dplyr::mutate(frozen_ffpe = uncurated4$is_ffpe) %>%
  dplyr::mutate(frozen_ffpe = dplyr::case_when(
    frozen_ffpe == "NO" ~ NA_character_,
    frozen_ffpe == "" ~ "FFPE",
    TRUE ~ NA_character_
  )) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated4$sample_type == "Metastatic" ~ "metastatic",
    uncurated4$sample_type == "Primary Tumor" ~ "primary",
    uncurated4$sample_type == "Solid Tissue Normal" ~ "normal"
  )) %>%
  dplyr::mutate(tissue_source = dplyr::case_when(
    uncurated4$initial_pathologic_diagnosis_method == "Core needle biopsy" ~ "biopsy",
    uncurated4$initial_pathologic_diagnosis_method == "Transurethral resection (TURBT)" ~ "TURP",
    TRUE ~ NA_character_,
  )) %>%
  dplyr::mutate(patient_id = gsub("-", ".", uncurated4$X_PATIENT)) %>%
  # dplyr::mutate(alt_sample_name = uncurated$OTHER_SAMPLE_ID) %>%
  dplyr::mutate(gleason_grade = uncurated4$gleason_score) %>%
  dplyr::mutate(gleason_major = uncurated4$primary_pattern) %>%
  dplyr::mutate(gleason_minor = uncurated4$secondary_pattern) %>%
  # dplyr::mutate(source_of_gleason = "biopsy") %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade == 6 ~ "<=6",
    gleason_major == 3 & gleason_minor == 4 ~ "3+4",
    gleason_major == 4 & gleason_minor == 3 ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8",
    TRUE ~ NA_character_
  )) %>%
  dplyr::mutate(zone_of_origin = uncurated4$zone_of_origin) %>%
  dplyr::mutate(zone_of_origin = dplyr::case_when(
    zone_of_origin == "Peripheral Zone" ~ "peripheral",
    zone_of_origin == "Overlapping / Multiple Zones" ~ "mixed",
    zone_of_origin == "Central Zone" ~ "central",
    zone_of_origin == "Transition Zone" ~ "transitional",
    zone_of_origin == "[Not Available]" ~ NA_character_,
    TRUE ~ NA_character_
  )) %>%
  # NOTE: Time was already in days, do not scale with days per month
  dplyr::mutate(year_diagnosis = uncurated4$year_of_initial_pathologic_diagnosis) %>%
  dplyr::mutate(overall_survival_status = uncurated4$OS) %>%
  # dplyr::mutate(days_to_overall_survival = as.numeric(uncurated4$OS.time) * 30.5) %>%
  dplyr::mutate(days_to_overall_survival = as.numeric(uncurated4$OS.time)) %>%
  # DSS = Disease Specific Survival, not recurrence! PF = Progression Free
  dplyr::mutate(days_to_disease_specific_recurrence = ifelse(
    is.na(uncurated4$days_to_first_biochemical_recurrence),
    # No known BCR date
    uncurated4$days_to_last_followup,
    # Known BCR date, taking first recurrence time
    uncurated4$days_to_first_biochemical_recurrence
  )) %>%
  dplyr::mutate(disease_specific_recurrence_status = as.integer(!is.na(uncurated4$"days_to_first_biochemical_recurrence"))) %>%
  # dplyr::mutate(days_to_disease_specific_recurrence = as.numeric(uncurated4$DSS.time)) %>%
  # dplyr::mutate(disease_specific_recurrence_status = uncurated4$DSS) %>%
  # dplyr::mutate(days_to_disease_specific_recurrence = as.numeric(uncurated4$DSS.time) * 30.5) %>%
  # dplyr::mutate(days_to_disease_specific_recurrence = as.numeric(uncurated4$DSS.time)) %>%
  # dplyr::mutate(disease_free_interval_status = uncurated$DFI) %>%
  # dplyr::mutate(days_to_disease_free_interval = as.numeric(uncurated$DFI.time) * 30.5) %>%
  # dplyr::mutate(progression_free_interval_status = uncurated$PFI) %>%
  # dplyr::mutate(days_to_progression_free_interval = as.numeric(uncurated$PFI.time) * 30.5) %>%
  dplyr::mutate(psa = uncurated4$psa_value) %>%
  dplyr::mutate(age_at_initial_diagnosis = uncurated4$age_at_initial_pathologic_diagnosis) %>%
  dplyr::mutate(M_stage = uncurated4$clinical_M) %>%
  dplyr::mutate(M_stage = dplyr::case_when(
    M_stage == "M0" ~ 0,
    M_stage %in% c("M1a", "M1b", "M1c") ~ 1,
    TRUE ~ NA_real_
  )) %>%
  # stringr:: commands return true NA not character NA
  dplyr::mutate(M_substage = stringr::str_sub(uncurated4$clinical_M, 3, 3)) %>%
  dplyr::mutate(T_clinical = as.integer(substr(uncurated4$clinical_T, 2, 2))) %>%
  dplyr::mutate(T_substage_clinical = stringr::str_extract(uncurated4$clinical_T, "[a-c]+")) %>%
  dplyr::mutate(T_pathological = as.integer(substr(uncurated4$pathologic_T, 2, 2))) %>%
  dplyr::mutate(T_substage_pathological = stringr::str_extract(uncurated4$pathologic_T, "[a-c]+")) %>%
  dplyr::mutate(therapy_radiation_initial = uncurated4$radiation_therapy) %>%
  # Radiation treatment given at initial treatment
  dplyr::mutate(therapy_radiation_initial = dplyr::case_when(
    therapy_radiation_initial == "YES" ~ 1,
    therapy_radiation_initial == "NO" ~ 0,
    therapy_radiation_initial == "" ~ NA_real_
  )) %>%
  dplyr::mutate(other_treatment = uncurated4$targeted_molecular_therapy) %>%
  # Add additional treatments
  dplyr::mutate(other_treatment = dplyr::case_when(
    other_treatment == "YES" ~ "targeted_molecular_therapy",
    other_treatment == "NO" ~ "",
    other_treatment == "" ~ NA_character_
  )) %>%
  # Nx, N0 or N1 if findings in lymph nodes
  dplyr::mutate(N_stage = uncurated4$pathologic_N) %>%
  dplyr::mutate(N_stage = dplyr::case_when(
    N_stage == "N0" ~ 0,
    N_stage == "N1" ~ 1,
    N_stage == "" ~ NA_real_
  )) %>%
  dplyr::mutate(seminal_vesicle_invasion = as.integer(
    uncurated4$diagnostic_ct_abd_pelvis_result == "Extraprostatic Extension  Localized (e.g. seminal vesicles)" |
      uncurated4$diagnostic_mri_result %in% c("Extraprostatic Extension Localized (e.g. seminal vesicles)", "Extraprostatic Extension Localized (e.g. seminal vesicles)|Extraprostatic Extension (regional lymphadenopathy) [e.g. cN1]")
  )) %>%
  dplyr::mutate(angiolymphatic_invasion = as.integer(
    uncurated4$diagnostic_ct_abd_pelvis_result == "Extraprostatic Extension (regional lymphadenopathy)[e.g. cN1]" |
      uncurated4$diagnostic_mri_result == "Extraprostatic Extension Localized (e.g. seminal vesicles)|Extraprostatic Extension (regional lymphadenopathy) [e.g. cN1]"
  ))

rownames(curated4) <- curated4$sample_name

# Append additional fields from the cgdsr
curated <- curated[rownames(curated4), ]
# Insert additional info where available from the cgdsr data extraction
for (colname in c("alt_sample_name", "ERG_fusion_GEX", "race", "tumor_purity_pathology", "tumor_purity_demixt", "tumor_purity_absolute", "AR_activity", "genome_altered")) {
  curated4[, colname] <- curated[, colname]
}

clinical_tcga <- curated4

save(clinical_tcga, file = "data-raw/clinical_tcga.RData")

###
#
# True et al. Proc Natl Acad Sci U S A 2006 Jul 18;103(29):10991-6.
# curated from GEO's clinical metadata (GSM)
#
###

# load series and platform data from GEO

gset <- GEOquery::getGEO("GSE5132", GSEMatrix = TRUE, getGPL = TRUE)

# for reasons unknown the clinical is split in a set of 31 samples and a set of 1 sample
uncurated1 <- Biobase::pData(gset[[1]])
uncurated2 <- Biobase::pData(gset[[2]])

curated1 <- initial_curated_df(
  df_rownames = rownames(uncurated1),
  template_name = "data-raw/template_prad.csv"
)

curated2 <- initial_curated_df(
  df_rownames = rownames(uncurated2),
  template_name = "data-raw/template_prad.csv"
)

#  additional issue is that, because the data seems to have been put in at random, most of it will have
# to be pulled 'manually' rather than algorithmically
curated1 <- curated1 %>%
  dplyr::mutate(study_name = "True et al.") %>%
  dplyr::mutate(sample_name = uncurated1$geo_accession) %>%
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(patient_id = unlist(lapply(stringr::str_split(uncurated1$title, "_"), function(x) x[2]))) %>%
  dplyr::mutate(gleason_major = as.numeric(substr(unlist(lapply(stringr::str_split(uncurated1$description, " "), function(x) x[9])), 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.numeric(substr(unlist(lapply(stringr::str_split(uncurated1$description, " "), function(x) x[9])), 3, 3))) %>%
  dplyr::mutate(gleason_grade = gleason_major + gleason_minor) %>%
  dplyr::mutate(microdissected = 1) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade %in% 4:6 ~ "<=6",
    gleason_major == 3 & gleason_minor == 4 ~ "3+4",
    gleason_major == 4 & gleason_minor == 3 ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8",
    TRUE ~ "NA"
  ))
curated1[1, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[1, "Prostate Cancer patient 03-138A Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[1, "tumor_margins_positive"] <- 0
curated1[1, "other_treatment"] <- substr(unlist(stringr::str_split(uncurated1[1, "Prostate Cancer patient 03-138A Gleason_Score:ch1"], " "))[8], 11, 20)
curated1[1, "therapy_radiation_initial"] <- 0
curated1[1, "therapy_radiation_salvage"] <- 0
curated1[1, "therapy_hormonal_initial"] <- 0
curated1[1, "other_feature"] <- paste(
  unlist(stringr::str_split(uncurated1[1, "Prostate Cancer patient 03-138A Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[1, "Prostate Cancer patient 03-138A Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[1, "Prostate Cancer patient 03-138A Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[1, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[1, "Prostate Cancer patient 03-138A Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[2, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[2, "Prostate Cancer patient 03-135C Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[2, "tumor_margins_positive"] <- 0
curated1[2, "therapy_radiation_initial"] <- 0
curated1[2, "therapy_radiation_salvage"] <- 0
curated1[2, "therapy_hormonal_initial"] <- 0
curated1[2, "other_feature"] <- paste(
  unlist(stringr::str_split(uncurated1[2, "Prostate Cancer patient 03-135C Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[2, "Prostate Cancer patient 03-135C Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[2, "Prostate Cancer patient 03-135C Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[2, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[2, "Prostate Cancer patient 03-135C Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[3, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[3, "Prostate Cancer patient 03-158F Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[3, "tumor_margins_positive"] <- 0
curated1[3, "therapy_radiation_initial"] <- 0
curated1[3, "therapy_radiation_salvage"] <- 0
curated1[3, "therapy_hormonal_initial"] <- 0
curated1[3, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[3, "Prostate Cancer patient 03-158F Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[3, "Prostate Cancer patient 03-158F Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[3, "Prostate Cancer patient 03-158F Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[3, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[3, "Prostate Cancer patient 03-158F Gleason_Score:ch2"], " "))[3], 17, 17))


curated1[4, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[4, " LCM_Gleason_Pattern:ch2"], " "))[4], 5, 8))
curated1[4, "tumor_margins_positive"] <- 0
curated1[4, "other_treatment"] <- paste(substr(unlist(stringr::str_split(uncurated1[4, " LCM_Gleason_Pattern:ch2"], " "))[7], 11, 20), unlist(stringr::str_split(uncurated1[4, " LCM_Gleason_Pattern:ch2"], " "))[8], sep = "_")
curated1[4, "therapy_radiation_initial"] <- 0
curated1[4, "therapy_radiation_salvage"] <- 0
curated1[4, "therapy_hormonal_initial"] <- 0
curated1[4, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[4, " LCM_Gleason_Pattern:ch2"], " "))[3],
  unlist(stringr::str_split(uncurated1[4, " LCM_Gleason_Pattern:ch2"], " "))[5],
  unlist(stringr::str_split(uncurated1[4, "characteristics_ch2.1"], " "))[2],
  sep = "|"
)
curated1[4, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[4, " LCM_Gleason_Pattern:ch2"], " "))[2], 17, 17))


curated1[5, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[5, " LCM_Gleason_Pattern:ch1"], " "))[4], 5, 8))
curated1[5, "tumor_margins_positive"] <- 0
curated1[5, "therapy_radiation_initial"] <- 0
curated1[5, "therapy_radiation_salvage"] <- 0
curated1[5, "therapy_hormonal_initial"] <- 0
curated1[5, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[5, " LCM_Gleason_Pattern:ch1"], " "))[3],
  unlist(stringr::str_split(uncurated1[5, " LCM_Gleason_Pattern:ch1"], " "))[5],
  unlist(stringr::str_split(uncurated1[5, "characteristics_ch1.1"], " "))[2],
  sep = "|"
)
curated1[6, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[5, " LCM_Gleason_Pattern:ch1"], " "))[2], 17, 17))


curated1[6, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[6, "Prostate Cancer patient 03-060A Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[6, "tumor_margins_positive"] <- 1
curated1[6, "therapy_radiation_initial"] <- 0
curated1[6, "therapy_radiation_salvage"] <- 0
curated1[6, "therapy_hormonal_initial"] <- 0
curated1[6, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[6, "Prostate Cancer patient 03-060A Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[6, "Prostate Cancer patient 03-060A Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[6, "Prostate Cancer patient 03-060A Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[6, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[6, "Prostate Cancer patient 03-060A Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[7, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[7, " LCM_Gleason_Pattern:ch2"], " "))[4], 5, 8))
curated1[7, "tumor_margins_positive"] <- 1
curated1[7, "therapy_radiation_initial"] <- 0
curated1[7, "therapy_radiation_salvage"] <- 0
curated1[7, "therapy_hormonal_initial"] <- 0
curated1[7, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[7, " LCM_Gleason_Pattern:ch2"], " "))[3],
  unlist(stringr::str_split(uncurated1[7, " LCM_Gleason_Pattern:ch2"], " "))[5],
  unlist(stringr::str_split(uncurated1[7, "characteristics_ch2.1"], " "))[2],
  sep = "|"
)
curated1[7, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[7, "characteristics_ch2.1"], " "))[3], 17, 17))

curated1[8, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[8, "Prostate Cancer patient 03-066C Gleason_Score:ch2"], " "))[8], 5, 8))
curated1[8, "tumor_margins_positive"] <- 1
curated1[8, "therapy_radiation_initial"] <- 0
curated1[8, "therapy_radiation_salvage"] <- 0
curated1[8, "therapy_hormonal_initial"] <- 0
curated1[8, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[8, "Prostate Cancer patient 03-066C Gleason_Score:ch2"], " "))[7],
  unlist(stringr::str_split(uncurated1[8, "Prostate Cancer patient 03-066C Gleason_Score:ch2"], " "))[9],
  unlist(stringr::str_split(uncurated1[8, "Prostate Cancer patient 03-066C Gleason_Score:ch2"], " "))[5],
  sep = "|"
)
curated1[8, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[8, "Prostate Cancer patient 03-066C Gleason_Score:ch2"], " "))[6], 17, 17))

curated1[9, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[9, "Prostate Cancer patient 03-159 Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[9, "tumor_margins_positive"] <- 1
curated1[9, "therapy_radiation_initial"] <- 0
curated1[9, "therapy_radiation_salvage"] <- 0
curated1[9, "therapy_hormonal_initial"] <- 0
curated1[9, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[9, "Prostate Cancer patient 03-159 Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[9, "Prostate Cancer patient 03-159 Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[9, "Prostate Cancer patient 03-159 Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[9, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[9, "Prostate Cancer patient 03-159 Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[10, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[10, "Prostate Cancer patient 03-021F Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[10, "tumor_margins_positive"] <- 1
curated1[10, "other_treatment"] <- substr(unlist(stringr::str_split(uncurated1[10, "Prostate Cancer patient 03-021F Gleason_Score:ch1"], " "))[8], 11, 20)
curated1[10, "therapy_radiation_initial"] <- 0
curated1[10, "therapy_radiation_salvage"] <- 0
curated1[10, "therapy_hormonal_initial"] <- 0
curated1[10, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[10, "Prostate Cancer patient 03-021F Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[10, "Prostate Cancer patient 03-021F Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[10, "Prostate Cancer patient 03-021F Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[10, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[10, "Prostate Cancer patient 03-021F Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[11, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[11, "Prostate Cancer patient 02-003E Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[11, "tumor_margins_positive"] <- 0
curated1[11, "therapy_radiation_initial"] <- 0
curated1[11, "therapy_radiation_salvage"] <- 0
curated1[11, "therapy_hormonal_initial"] <- 0
curated1[11, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[11, "Prostate Cancer patient 02-003E Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[11, "Prostate Cancer patient 02-003E Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[11, "Prostate Cancer patient 02-003E Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[11, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[11, "Prostate Cancer patient 02-003E Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[12, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[8], 5, 8))
curated1[12, "tumor_margins_positive"] <- 0
curated1[12, "other_treatment"] <- paste(substr(unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[11], 11, 15), unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[12], sep = "_")
curated1[12, "therapy_radiation_initial"] <- 0
curated1[12, "therapy_radiation_salvage"] <- 0
curated1[12, "therapy_hormonal_initial"] <- 0
curated1[12, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[7],
  unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[9],
  unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[5],
  sep = "|"
)
curated1[12, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[12, "Prostate Cancer patient 03-207C Gleason_Score:ch1"], " "))[6], 17, 17))

curated1[13, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[13, "LCM_Gleason_Pattern:ch2"], " "))[4], 5, 8))
curated1[13, "tumor_margins_positive"] <- 1
curated1[13, "therapy_radiation_initial"] <- 0
curated1[13, "therapy_radiation_salvage"] <- 0
curated1[13, "therapy_hormonal_initial"] <- 0
curated1[13, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[13, "LCM_Gleason_Pattern:ch2"], " "))[3],
  unlist(stringr::str_split(uncurated1[13, "LCM_Gleason_Pattern:ch2"], " "))[5],
  unlist(stringr::str_split(uncurated1[13, "characteristics_ch2.1"], " "))[1],
  sep = "|"
)
curated1[13, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[13, "LCM_Gleason_Pattern:ch2"], " "))[2], 17, 17))

curated1[14, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[14, " LCM_Gleason_Pattern:ch2"], " "))[4], 5, 8))
curated1[14, "tumor_margins_positive"] <- 0
curated1[14, "therapy_radiation_initial"] <- 0
curated1[14, "therapy_radiation_salvage"] <- 0
curated1[14, "therapy_hormonal_initial"] <- 0
curated1[14, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[14, " LCM_Gleason_Pattern:ch2"], " "))[3],
  unlist(stringr::str_split(uncurated1[14, " LCM_Gleason_Pattern:ch2"], " "))[5],
  unlist(stringr::str_split(uncurated1[14, "characteristics_ch2.1"], " "))[2],
  sep = "|"
)
curated1[14, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[14, " LCM_Gleason_Pattern:ch2"], " "))[2], 17, 17))

curated1[15, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[15, " LCM_Gleason_Pattern:ch1"], " "))[4], 5, 8))
curated1[15, "tumor_margins_positive"] <- 1
curated1[15, "therapy_radiation_initial"] <- 0
curated1[15, "therapy_radiation_salvage"] <- 0
curated1[15, "therapy_hormonal_initial"] <- 0
curated1[15, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[15, " LCM_Gleason_Pattern:ch1"], " "))[3],
  unlist(stringr::str_split(uncurated1[15, " LCM_Gleason_Pattern:ch1"], " "))[5],
  unlist(stringr::str_split(uncurated1[15, "characteristics_ch1.1"], " "))[2],
  sep = "|"
)
curated1[15, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[15, " LCM_Gleason_Pattern:ch1"], " "))[2], 17, 17))

curated1[16, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[16, "Prostate Cancer patient 03-068D Gleason_Score:ch2"], " "))[8], 5, 8))
curated1[16, "tumor_margins_positive"] <- 1
curated1[16, "therapy_radiation_initial"] <- 0
curated1[16, "therapy_radiation_salvage"] <- 0
curated1[16, "therapy_hormonal_initial"] <- 0
curated1[16, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[16, "Prostate Cancer patient 03-068D Gleason_Score:ch2"], " "))[7],
  unlist(stringr::str_split(uncurated1[16, "Prostate Cancer patient 03-068D Gleason_Score:ch2"], " "))[9],
  unlist(stringr::str_split(uncurated1[16, "Prostate Cancer patient 03-068D Gleason_Score:ch2"], " "))[5],
  sep = "|"
)
curated1[16, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[16, "Prostate Cancer patient 03-068D Gleason_Score:ch2"], " "))[6], 17, 17))

curated1[17, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[17, "Prostate Cancer patient 03-141C Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[17, "tumor_margins_positive"] <- 1
curated1[17, "therapy_radiation_initial"] <- 0
curated1[17, "therapy_radiation_salvage"] <- 0
curated1[17, "therapy_hormonal_initial"] <- 0
curated1[17, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[17, "Prostate Cancer patient 03-141C Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[17, "Prostate Cancer patient 03-141C Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[17, "Prostate Cancer patient 03-141C Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[17, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[17, "Prostate Cancer patient 03-141C Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[18, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[18, "Prostate Cancer patient 02-053C Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[18, "tumor_margins_positive"] <- 0
curated1[18, "therapy_radiation_initial"] <- 0
curated1[18, "therapy_radiation_salvage"] <- 0
curated1[18, "therapy_hormonal_initial"] <- 0
curated1[18, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[18, "Prostate Cancer patient 02-053C Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[18, "Prostate Cancer patient 02-053C Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[18, "Prostate Cancer patient 02-053C Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[18, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[18, "Prostate Cancer patient 02-053C Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[19, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[19, "Prostate Cancer patient 04-030A Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[19, "tumor_margins_positive"] <- 0
curated1[19, "therapy_radiation_initial"] <- 0
curated1[19, "therapy_radiation_salvage"] <- 0
curated1[19, "therapy_hormonal_initial"] <- 0
curated1[19, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[19, "Prostate Cancer patient 04-030A Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[19, "Prostate Cancer patient 04-030A Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[19, "Prostate Cancer patient 04-030A Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[19, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[19, "Prostate Cancer patient 04-030A Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[20, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[20, " LCM_Gleason_Pattern:ch1"], " "))[4], 5, 8))
curated1[20, "tumor_margins_positive"] <- 1
curated1[20, "therapy_radiation_initial"] <- 0
curated1[20, "therapy_radiation_salvage"] <- 0
curated1[20, "therapy_hormonal_initial"] <- 0
curated1[20, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[20, " LCM_Gleason_Pattern:ch1"], " "))[3],
  unlist(stringr::str_split(uncurated1[20, " LCM_Gleason_Pattern:ch1"], " "))[5],
  unlist(stringr::str_split(uncurated1[20, "characteristics_ch1.1"], " "))[2],
  sep = "|"
)
curated1[20, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[20, " LCM_Gleason_Pattern:ch1"], " "))[2], 17, 17))

curated1[21, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[21, "LCM_Gleason_Pattern:ch2"], " "))[4], 5, 8))
curated1[21, "tumor_margins_positive"] <- 0
curated1[21, "other_treatment"] <- substr(unlist(stringr::str_split(uncurated1[21, "LCM_Gleason_Pattern:ch2"], " "))[7], 11, 20)
curated1[21, "therapy_radiation_initial"] <- 0
curated1[21, "therapy_radiation_salvage"] <- 0
curated1[21, "therapy_hormonal_initial"] <- 0
curated1[21, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[21, "LCM_Gleason_Pattern:ch2"], " "))[3],
  unlist(stringr::str_split(uncurated1[21, "LCM_Gleason_Pattern:ch2"], " "))[5],
  unlist(stringr::str_split(uncurated1[21, "characteristics_ch2.1"], " "))[1],
  sep = "|"
)
curated1[21, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[21, "LCM_Gleason_Pattern:ch2"], " "))[2], 17, 17))

curated1[22, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[22, "Prostate Cancer patient 03-184C Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[22, "tumor_margins_positive"] <- 0
curated1[22, "therapy_radiation_initial"] <- 0
curated1[22, "therapy_radiation_salvage"] <- 0
curated1[22, "therapy_hormonal_initial"] <- 0
curated1[22, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[22, "Prostate Cancer patient 03-184C Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[22, "Prostate Cancer patient 03-184C Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[22, "Prostate Cancer patient 03-184C Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[22, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[22, "Prostate Cancer patient 03-184C Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[23, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[23, "tumor_margins_positive"] <- 1
curated1[23, "other_treatment"] <- paste(substr(unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[8], 11, 20), unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[9], sep = "_")
curated1[23, "therapy_radiation_initial"] <- 0
curated1[23, "therapy_radiation_salvage"] <- 0
curated1[23, "therapy_hormonal_initial"] <- 0
curated1[23, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[23, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[23, "Prostate Cancer patient 03-152A Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[24, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[24, "Prostate Cancer patient 03-029B Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[24, "tumor_margins_positive"] <- 0
curated1[24, "therapy_radiation_initial"] <- 0
curated1[24, "therapy_radiation_salvage"] <- 0
curated1[24, "therapy_hormonal_initial"] <- 0
curated1[24, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[24, "Prostate Cancer patient 03-029B Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[24, "Prostate Cancer patient 03-029B Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[24, "Prostate Cancer patient 03-029B Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[24, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[24, "Prostate Cancer patient 03-029B Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[25, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[25, "Prostate Cancer patient 03-015F Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[25, "tumor_margins_positive"] <- 1
curated1[25, "therapy_radiation_initial"] <- 0
curated1[25, "therapy_radiation_salvage"] <- 0
curated1[25, "therapy_hormonal_initial"] <- 0
curated1[25, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[25, "Prostate Cancer patient 03-015F Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[25, "Prostate Cancer patient 03-015F Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[25, "Prostate Cancer patient 03-015F Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[25, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[25, "Prostate Cancer patient 03-015F Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[26, "psa"] <- NA
curated1[26, "tumor_margins_positive"] <- 1
curated1[26, "therapy_radiation_initial"] <- 0
curated1[26, "therapy_radiation_salvage"] <- 0
curated1[26, "therapy_hormonal_initial"] <- 0
curated1[26, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[26, "Prostate Cancer patient 03-119D Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[26, "Prostate Cancer patient 03-119D Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[26, "Prostate Cancer patient 03-119D Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[26, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[26, "Prostate Cancer patient 03-119D Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[27, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[27, "tumor_margins_positive"] <- 0
curated1[27, "other_treatment"] <- paste(substr(unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[8], 11, 20), unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[9], sep = "_")
curated1[27, "therapy_radiation_initial"] <- 0
curated1[27, "therapy_radiation_salvage"] <- 0
curated1[27, "therapy_hormonal_initial"] <- 0
curated1[27, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[27, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[27, "Prostate Cancer patient 03-140B Gleason_Score:ch1"], " "))[3], 17, 17))

curated1[28, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[28, "LCM_Gleason_Pattern:ch1"], " "))[4], 5, 8))
curated1[28, "tumor_margins_positive"] <- 0
curated1[28, "therapy_radiation_initial"] <- 0
curated1[28, "therapy_radiation_salvage"] <- 0
curated1[28, "therapy_hormonal_initial"] <- 0
curated1[28, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[28, "LCM_Gleason_Pattern:ch1"], " "))[3],
  unlist(stringr::str_split(uncurated1[28, "LCM_Gleason_Pattern:ch1"], " "))[5],
  unlist(stringr::str_split(uncurated1[28, "characteristics_ch1.1"], " "))[1],
  sep = "|"
)
curated1[28, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[28, "LCM_Gleason_Pattern:ch1"], " "))[2], 17, 17))

curated1[29, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[29, " LCM_Gleason_Pattern:ch2"], " "))[4], 5, 8))
curated1[29, "tumor_margins_positive"] <- 0
curated1[29, "therapy_radiation_initial"] <- 0
curated1[29, "therapy_radiation_salvage"] <- 0
curated1[29, "therapy_hormonal_initial"] <- 0
curated1[29, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[29, " LCM_Gleason_Pattern:ch2"], " "))[3],
  unlist(stringr::str_split(uncurated1[29, " LCM_Gleason_Pattern:ch2"], " "))[5],
  unlist(stringr::str_split(uncurated1[29, "characteristics_ch2.1"], " "))[2],
  sep = "|"
)
curated1[29, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[29, " LCM_Gleason_Pattern:ch2"], " "))[2], 17, 17))

curated1[30, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[30, "Prostate Cancer patient 03-063A Gleason_Score:ch2"], " "))[5], 5, 8))
curated1[30, "tumor_margins_positive"] <- 0
curated1[30, "therapy_radiation_initial"] <- 0
curated1[30, "therapy_radiation_salvage"] <- 0
curated1[30, "therapy_hormonal_initial"] <- 0
curated1[30, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[30, "Prostate Cancer patient 03-063A Gleason_Score:ch2"], " "))[4],
  unlist(stringr::str_split(uncurated1[30, "Prostate Cancer patient 03-063A Gleason_Score:ch2"], " "))[6],
  unlist(stringr::str_split(uncurated1[30, "Prostate Cancer patient 03-063A Gleason_Score:ch2"], " "))[2],
  sep = "|"
)
curated1[30, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[30, "Prostate Cancer patient 03-063A Gleason_Score:ch2"], " "))[3], 17, 17))

curated1[31, "psa"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[5], 5, 8))
curated1[31, "tumor_margins_positive"] <- 0
curated1[31, "other_treatment"] <- paste(substr(unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[8], 11, 20), unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[9], sep = "_")
curated1[31, "therapy_radiation_initial"] <- 0
curated1[31, "therapy_radiation_salvage"] <- 0
curated1[31, "therapy_hormonal_initial"] <- 0
curated1[31, "other_feature"] <- paste(unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[4],
  unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[6],
  unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[2],
  sep = "|"
)
curated1[31, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated1[31, "Prostate Cancer patient 03-055H Gleason_Score:ch1"], " "))[3], 17, 17))

###########################
###########################

# being only one line the second dataset does not cause such inconvenients

curated2 <- curated2 %>%
  dplyr::mutate(study_name = "True et al.") %>%
  dplyr::mutate(sample_name = uncurated2$geo_accession) %>%
  dplyr::mutate(sample_paired = 1) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(patient_id = unlist(lapply(stringr::str_split(uncurated2$title, "_"), function(x) x[2]))) %>%
  dplyr::mutate(gleason_major = as.numeric(substr(unlist(lapply(stringr::str_split(uncurated2$description, " "), function(x) x[9])), 1, 1))) %>%
  dplyr::mutate(gleason_minor = as.numeric(substr(unlist(lapply(stringr::str_split(uncurated2$description, " "), function(x) x[9])), 3, 3))) %>%
  dplyr::mutate(gleason_grade = gleason_major + gleason_minor) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade %in% 4:6 ~ "<=6",
    gleason_major == 3 & gleason_minor == 4 ~ "3+4",
    gleason_major == 4 & gleason_minor == 3 ~ "4+3",
    gleason_grade %in% 8:10 ~ ">=8",
    TRUE ~ "NA"
  )) %>%
  dplyr::mutate(psa = as.numeric(substr(unlist(stringr::str_split(uncurated2$characteristics_ch1, " "))[13], 5, 8))) %>%
  dplyr::mutate(tumor_margins_positive = 0) %>%
  dplyr::mutate(therapy_radiation_initial = 0) %>%
  dplyr::mutate(therapy_radiation_salvage = 0) %>%
  dplyr::mutate(therapy_hormonal_initial = 0) %>%
  dplyr::mutate(microdissected = 1) %>%
  dplyr::mutate(other_feature = paste(
    unlist(stringr::str_split(uncurated2$characteristics_ch1, " "))[12],
    unlist(stringr::str_split(uncurated2$characteristics_ch1, " "))[14],
    unlist(stringr::str_split(uncurated2$characteristics_ch1, " "))[10],
    paste("tertiary_pattern", stringr::str_split(uncurated2$characteristics_ch1, " ")[[1]][8], sep = ":"),
    sep = "|"
  ))

curated2[1, "grade_group"] <- as.numeric(substr(unlist(stringr::str_split(uncurated2[1, "Prostate Cancer patient 02-209C Gleason_Score:ch1"], " "))[7], 17, 17))

# the datastes are merged in a way that preserves the order of the GEO sample indexes

clinical_true <- rbind(curated1[1:10, ], curated2[1, ], curated1[11:31, ])
# Fix to grade groups
clinical_true[which(clinical_true[, "gleason_grade"] <= 6), "grade_group"] <- "<=6"
clinical_true[which(clinical_true[, "gleason_major"] == 3 & clinical_true[, "gleason_minor"] == 4), "grade_group"] <- "3+4"
clinical_true[which(clinical_true[, "gleason_major"] == 4 & clinical_true[, "gleason_minor"] == 3), "grade_group"] <- "4+3"
clinical_true[which(clinical_true[, "gleason_grade"] >= 8), "grade_group"] <- ">=8"


save(clinical_true, file = "./data-raw//clinical_true.RData")

##################################################################
#
# Wallace et al.
# Curate clinical data from GEO's clinical metadata fields (GSM)
#
##################################################################

gset <- GEOquery::getGEO("GSE6956", GSEMatrix = TRUE, getGPL = FALSE)

uncurated <- Biobase::pData(gset[[1]])

unmatched_healty_tissue <- c("GSM160418", "GSM160419", "GSM160420", "GSM160421", "GSM160422", "GSM160430")

uncurated <- uncurated[!is.element(uncurated$geo_accession, unmatched_healty_tissue), ]

# Base curation metadat
curated <- initial_curated_internal(df_rownames = rownames(uncurated))

curated <- curated %>%
  dplyr::mutate(study_name = "Wallace et al.") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(patient_id = stringr::str_extract(uncurated$title, "patient \\d+")) %>%
  dplyr::mutate(alt_sample_name = stringr::str_extract(uncurated$title, "patient \\d+")) %>%
  dplyr::mutate(gleason_grade = dplyr::case_when(
    uncurated$"gleason sum:ch1" == "NA" ~ NA_character_,
    TRUE ~ uncurated$"gleason sum:ch1"
  )) %>%
  dplyr::mutate(gleason_grade = as.numeric(gleason_grade)) %>%
  dplyr::mutate(race = dplyr::case_when(
    uncurated$"race:ch1" == "African American" ~ "african_american",
    uncurated$"race:ch1" == "Caucasian" ~ "caucasian",
    uncurated$"race:ch1" == "NA" ~ NA_character_
  )) %>%
  dplyr::mutate(smoking_status = as.numeric(dplyr::case_when(
    uncurated$"smoking status:ch1" == "Current" ~ "1",
    uncurated$"smoking status:ch1" == "Past" ~ "1",
    uncurated$"smoking status:ch1" == "Never" ~ "0",
    uncurated$"smoking status:ch1" == "Unknown" ~ NA_character_,
    uncurated$"smoking status:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(angiolymphatic_invasion = as.numeric(dplyr::case_when(
    uncurated$"angio lymphatic invasion:ch1" == "Yes" ~ "1",
    uncurated$"angio lymphatic invasion:ch1" == "No" ~ "0",
    uncurated$"angio lymphatic invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(seminal_vesicle_invasion = as.numeric(dplyr::case_when(
    uncurated$"seminal vesicle invasion:ch1" == "Yes" ~ "1",
    uncurated$"seminal vesicle invasion:ch1" == "No" ~ "0",
    uncurated$"seminal vesicle invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(perineural_invasion = as.numeric(dplyr::case_when(
    uncurated$"perineural invasion:ch1" == "Yes" ~ "1",
    uncurated$"perineural invasion:ch1" == "No" ~ "0",
    uncurated$"perineural invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(extraprostatic_extension = as.numeric(dplyr::case_when(
    uncurated$"extraprostatic extension:ch1" == "Focal" ~ "1",
    uncurated$"extraprostatic extension:ch1" == "Multifocal" ~ "1",
    uncurated$"extraprostatic extension:ch1" == "Established" ~ "1",
    uncurated$"extraprostatic extension:ch1" == "None" ~ "0",
    uncurated$"perineural invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(tumor_margins_positive = as.numeric(dplyr::case_when(
    uncurated$"are surgical margins involved:ch1" == "Tumor focal at margin" ~ "1",
    uncurated$"are surgical margins involved:ch1" == "Tumor widespread a surgical margins" ~ "1",
    uncurated$"are surgical margins involved:ch1" == "Tumor widespread at margin" ~ "1",
    uncurated$"are surgical margins involved:ch1" == "All surgical margins are free of tumor" ~ "0",
    uncurated$"are surgical margins involved:ch1" == "Unknown" ~ NA_character_,
    uncurated$"are surgical margins involved:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated$source_name_ch1 == "Adenocarcinoma (NOS) of the prostate" ~ "primary",
    uncurated$source_name_ch1 == "Normal prostate" ~ "normal"
  )) %>%
  dplyr::mutate(frozen_ffpe = "frozen") %>%
  dplyr::mutate(microdissected = 0) %>%
  dplyr::mutate(therapy_radiation_initial = 0) %>%
  dplyr::mutate(therapy_radiation_salvage = 0) %>%
  dplyr::mutate(therapy_surgery_initial = 0) %>%
  dplyr::mutate(therapy_hormonal_initial = 0)




clinical_wallace <- curated

save(clinical_wallace, file = "data-raw/clinical_wallace.RData")

###
#
# Wallace et al. alternative
#
###

# load series and platform data from GEO
gset <- GEOquery::getGEO("GSE6956", GSEMatrix = TRUE, getGPL = FALSE)

# clinical
uncurated <- Biobase::pData(gset[[1]])

# TDL: Retain also the non-matched normal samples, just indicate them with sample_matched-field
# unmatched_healty_tissue = c('GSM160418', 'GSM160419', 'GSM160420', 'GSM160421', 'GSM160422', 'GSM160430') # as determined by the GSE6956 metadata
# uncurated = uncurated[!is.element(uncurated$geo_accession, unmatched_healty_tissue), ]

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)


curated <- curated %>%
  dplyr::mutate(study_name = "Wallace et al.") %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(patient_id = stringr::str_extract(uncurated$title, "patient \\d+")) %>%
  dplyr::mutate(alt_sample_name = stringr::str_extract(uncurated$title, "patient \\d+")) %>%
  dplyr::mutate(gleason_grade = dplyr::case_when(
    uncurated$"gleason sum:ch1" == "NA" ~ NA_character_,
    TRUE ~ uncurated$"gleason sum:ch1"
  )) %>%
  dplyr::mutate(gleason_grade = as.numeric(gleason_grade)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    gleason_grade <= 6 ~ "<=6",
    gleason_grade == 7 ~ "7",
    gleason_grade >= 8 ~ ">=8",
  )) %>%
  dplyr::mutate(race = dplyr::case_when(
    uncurated$"race:ch1" == "African American" ~ "african_american",
    uncurated$"race:ch1" == "Caucasian" ~ "caucasian",
    uncurated$"race:ch1" == "NA" ~ NA_character_
  )) %>%
  dplyr::mutate(smoking_status = as.numeric(dplyr::case_when(
    uncurated$"smoking status:ch1" == "Current" ~ "1",
    uncurated$"smoking status:ch1" == "Past" ~ "1",
    uncurated$"smoking status:ch1" == "Never" ~ "0",
    uncurated$"smoking status:ch1" == "Unknown" ~ NA_character_,
    uncurated$"smoking status:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(angiolymphatic_invasion = as.numeric(dplyr::case_when(
    uncurated$"angio lymphatic invasion:ch1" == "Yes" ~ "1",
    uncurated$"angio lymphatic invasion:ch1" == "No" ~ "0",
    uncurated$"angio lymphatic invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(seminal_vesicle_invasion = as.numeric(dplyr::case_when(
    uncurated$"seminal vesicle invasion:ch1" == "Yes" ~ "1",
    uncurated$"seminal vesicle invasion:ch1" == "No" ~ "0",
    uncurated$"seminal vesicle invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(perineural_invasion = as.numeric(dplyr::case_when(
    uncurated$"perineural invasion:ch1" == "Yes" ~ "1",
    uncurated$"perineural invasion:ch1" == "No" ~ "0",
    uncurated$"perineural invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(extraprostatic_extension = as.numeric(dplyr::case_when(
    uncurated$"extraprostatic extension:ch1" == "Focal" ~ "1",
    uncurated$"extraprostatic extension:ch1" == "Multifocal" ~ "1",
    uncurated$"extraprostatic extension:ch1" == "Established" ~ "1",
    uncurated$"extraprostatic extension:ch1" == "None" ~ "0",
    uncurated$"perineural invasion:ch1" == "NA" ~ NA_character_
  ))) %>%
  dplyr::mutate(tumor_margins_positive = as.numeric(dplyr::case_when(
    uncurated$"are surgical margins involved:ch1" == "Tumor focal at margin" ~ "1",
    uncurated$"are surgical margins involved:ch1" == "Tumor widespread a surgical margins" ~ "1",
    uncurated$"are surgical margins involved:ch1" == "Tumor widespread at margin" ~ "1",
    uncurated$"are surgical margins involved:ch1" == "All surgical margins are free of tumor" ~ "0",
    uncurated$"are surgical margins involved:ch1" == "Unknown" ~ NA_character_,
    uncurated$"are surgical margins involved:ch1" == "NA" ~ NA_character_
  ))) %>%
  # The generic normals are not paired, adjacents were paired
  # dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(sample_paired = ifelse(substr(uncurated$title, 1, 6) == "Pooled", 0, 1)) %>%
  dplyr::mutate(sample_type = dplyr::case_when(
    uncurated$source_name_ch1 == "Adenocarcinoma (NOS) of the prostate" ~ "primary",
    # uncurated$source_name_ch1 == "Normal prostate" ~ 'adjacentnormal'
    uncurated$source_name_ch1 == "Normal prostate" ~ "normal"
  )) %>%
  dplyr::mutate(frozen_ffpe = "frozen") %>%
  dplyr::mutate(microdissected = 0) %>%
  # Pathological T stage was available without substage
  dplyr::mutate(T_pathological = as.numeric(uncurated$"pt stage:ch1")) %>%
  # "All tumors were resected adenocarcinomas that had not received any therapy prior to prostatectomy."
  dplyr::mutate(therapy_radiation_initial = 0) %>%
  dplyr::mutate(therapy_radiation_salvage = 0) %>%
  dplyr::mutate(therapy_surgery_initial = 0) %>%
  dplyr::mutate(therapy_hormonal_initial = 0)

# The unnamed matched normals are given unique identifiers
curated[which(is.na(curated$patient_id)), "patient_id"] <- paste0("unmatched normal ", 1:6)

clinical_wallace <- curated

save(clinical_wallace, file = "data-raw/clinical_wallace.RData")


###
#
# Wang et al.
# Source: GEO
#
###

gse <- GEOquery::getGEO("GSE8218", GSEMatrix = TRUE)

uncurated <- Biobase::pData(gse[[1]])

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

# Notable heterogeneity with percentages, and original publication created a multivariate classifier for purity; samples as best described by the dominating subtype
# Allow categorized labels although exact percentages would be optimal to use
# Following e.g.
# - https://www.pathologyoutlines.com/topic/prostateatrophy.html
# - https://www.europeanurology.com/article/S0302-2838(06)00700-7/fulltext
# - https://www.sciencedirect.com/science/article/pii/S0031302520309296
#
# ... defining atrophic glands as their own sample type as a confounder-like variable
curated <- curated %>%
  dplyr::mutate(study_name = "Wang et al.") %>%
  dplyr::mutate(patient_id = rownames(uncurated)) %>%
  dplyr::mutate(sample_name = row.names(uncurated)) %>%
  dplyr::mutate(alt_sample_name = uncurated$"title") %>%
  dplyr::mutate(sample_paired = 0) %>%
  ## TDL: This is not true for the sample, study, and its key aims
  # dplyr::mutate(sample_type = "primary") %>%
  ## TDL: Tumor purity of percentage is a key field rather than specific subsets
  # From GEO description: "The percentage of different cell types vary considerably among samples and were determined by pathologist."
  # Despite reported as 'prostate cancer sample(s)' in GEO, the samples are quite mixed.

  dplyr::mutate(sample_type = c("primary", "atrophic", "BPH", "stroma")[apply(cbind(
    "primary" = as.numeric(gsub("%", "", uncurated$"Percentage of Tumor:ch1")) / 100,
    "atrophic" = as.numeric(gsub("%", "", uncurated$"Percentage of Atrophic Gland:ch1")) / 100,
    "BPH" = as.numeric(gsub("%", "", uncurated$"Percentage of BPH:ch1")) / 100,
    "stroma" = as.numeric(gsub("%", "", uncurated$"Percentage of stroma:ch1")) / 100
  ), MARGIN = 1, FUN = function(x) {
    ifelse(all(x == "not Known"), NA_integer_, which.max(x))
  })]) %>%
  dplyr::mutate(tumor_purity_pathology = as.numeric(gsub("%", "", uncurated$"Percentage of Tumor:ch1")) / 100) %>%
  # dplyr::mutate(Percentage_of_Atrophic_Gland = uncurated$`Percentage of Atrophic Gland:ch1`) %>%
  # dplyr::mutate(Percentage_of_BPH = uncurated$`Percentage of BPH:ch1`) %>%
  # dplyr::mutate(Percentage_of_Stroma = uncurated$`Percentage of Stroma:ch1`) %>%
  # dplyr::mutate(Percentage_of_Tumor = uncurated$`Percentage of Tumor:ch1`)
  dplyr::mutate(other_sample = apply(cbind(
    # All histology subtypes were reported with precision of 2 (a percentage unit)
    "primary" = paste0("Perc_Tumor=", round(as.numeric(gsub("%", "", uncurated$"Percentage of Tumor:ch1")) / 100, 2)),
    "atrophic" = paste0("Perc_AtrophicGland=", round(as.numeric(gsub("%", "", uncurated$"Percentage of Atrophic Gland:ch1")) / 100, 2)),
    "BPH" = paste0("Perc_BPH=", round(as.numeric(gsub("%", "", uncurated$"Percentage of BPH:ch1")) / 100, 2)),
    "stroma" = paste0("Perc_Stroma=", round(as.numeric(gsub("%", "", uncurated$"Percentage of BPH:ch1")) / 100, 2))
  ), MARGIN = 1, FUN = function(x) {
    ifelse(all(x == "not Known"), NA_character_, paste0(x, collapse = "|"))
  }))

# Tabulated, the sample types discretized based on the dominant histological subtype:
# atrophic      BPH  primary
#      21       55       60

clinical_wang <- curated

save(clinical_wang, file = "data-raw/clinical_wang.RData")

###
#
# Weiner et al.
# Source: GEO
#
###

gset <- getGEO("GSE157548", GSEMatrix = TRUE, getGPL = TRUE)

uncurated <- Biobase::pData(gset[[1]])

curated <- initial_curated_internal(
  df_rownames = rownames(uncurated)
)

curated <- curated %>%
  dplyr::mutate(study_name = "Weiner et al.") %>%
  dplyr::mutate(sample_name = uncurated$geo_accession) %>%
  dplyr::mutate(patient_id = uncurated$geo_accession) %>%
  dplyr::mutate(race = dplyr::case_when(
    is.na(uncurated$"race:ch1") ~ NA_character_,
    uncurated$"race:ch1" == "Black" ~ "african_american",
    uncurated$"race:ch1" == "White" ~ "caucasian"
  )) %>%
  dplyr::mutate(psa = round(as.numeric(uncurated$"psa:ch1"), 2)) %>%
  dplyr::mutate(tissue_source = "prostatectomy") %>%
  dplyr::mutate(age_at_initial_diagnosis = dplyr::case_when(
    is.na(uncurated$"age:ch1") ~ NA_character_,
    TRUE ~ uncurated$"age:ch1"
  )) %>%
  dplyr::mutate(age_at_initial_diagnosis = as.numeric(age_at_initial_diagnosis)) %>%
  dplyr::mutate(grade_group = dplyr::case_when(
    uncurated$"grade group:ch1" == "1" ~ "<=6",
    uncurated$"grade group:ch1" == "2" ~ "3+4",
    uncurated$"grade group:ch1" == "3" ~ "4+3",
    uncurated$"grade group:ch1" == "4" ~ ">=8",
    uncurated$"grade group:ch1" == "5" ~ ">=8",
  )) %>%
  dplyr::mutate(frozen_ffpe = "FFPE") %>%
  dplyr::mutate(batch = dplyr::case_when(
    is.na(uncurated$"race:ch1") ~ "Durham Veterans Affairs Hospital",
    uncurated$"race:ch1" == "Black" ~ "Johns Hopkins Medical Institute",
    uncurated$"race:ch1" == "White" ~ "Johns Hopkins Medical Institute"
  )) %>%
  dplyr::mutate(therapy_radiation_initial = 0) %>%
  dplyr::mutate(therapy_radiation_salvage = 0) %>%
  dplyr::mutate(therapy_surgery_initial = 0) %>%
  dplyr::mutate(therapy_hormonal_initial = 0) %>%
  dplyr::mutate(sample_paired = 0) %>%
  dplyr::mutate(sample_type = "primary") %>%
  dplyr::mutate(microdissected = 0) %>%
  dplyr::mutate(extraprostatic_extension = dplyr::case_when(
    uncurated$"non-organ confined:ch1" == "Yes" ~ 1,
    uncurated$"non-organ confined:ch1" == "No" ~ 0
  )) %>%
  dplyr::mutate(
    other_sample =
      apply(
        cbind(
          paste0("grade_group=", uncurated$"grade group:ch1"),
          paste0("ifng_signature=", uncurated$"ifng signature:ch1"),
          paste0("igg_production=", uncurated$"igg production:ch1"),
          paste0("inflammation_signature=", uncurated$"inflammation signature:ch1"),
          paste0("nk_activity=", uncurated$"nk activity:ch1"),
          # Typo by original authors, concatenating the two vectors together based on which vector has NA-value
          paste0("plasma_cell_content=", ifelse(is.na(uncurated$"plasma cell content:ch1"), uncurated$"palsma cell content:ch1", uncurated$"plasma cell content:ch1"))
        ),
        MARGIN = 1, FUN = function(x) {
          paste(x, collapse = "|")
        }
      )
  )

clinical_weiner <- curated

save(clinical_weiner, file = "data-raw/clinical_weiner.RData")