p3/preprocess/Pancreatic_Cancer/code/TCGA.py

# Path Configuration
from tools.preprocess import *

# Processing context
trait = "Pancreatic_Cancer"

# Input paths
tcga_root_dir = "../DATA/TCGA"

# Output paths
out_data_file = "./output/preprocess/3/Pancreatic_Cancer/TCGA.csv"
out_gene_data_file = "./output/preprocess/3/Pancreatic_Cancer/gene_data/TCGA.csv"
out_clinical_data_file = "./output/preprocess/3/Pancreatic_Cancer/clinical_data/TCGA.csv"
json_path = "./output/preprocess/3/Pancreatic_Cancer/cohort_info.json"

# 1. Select the relevant subdirectory for pancreatic cancer
subdirectory = 'TCGA_Pancreatic_Cancer_(PAAD)'
cohort_dir = os.path.join(tcga_root_dir, subdirectory)

# 2. Get the file paths
clinical_file_path, genetic_file_path = tcga_get_relevant_filepaths(cohort_dir)

# 3. Load the data files  
clinical_df = pd.read_csv(clinical_file_path, index_col=0, sep='\t')
genetic_df = pd.read_csv(genetic_file_path, index_col=0, sep='\t')

# 4. Print clinical data columns
print("Clinical data columns:")
print(clinical_df.columns.tolist())
# Identify candidate columns
candidate_age_cols = ["age_at_initial_pathologic_diagnosis", "days_to_birth"]
candidate_gender_cols = ["gender"]

# Create preview dictionary directly from the columns list
# Since we cannot access the data files currently, we output the candidate column names
age_preview = {"Selected age columns": candidate_age_cols}
gender_preview = {"Selected gender columns": candidate_gender_cols}

print("Age columns:")
print(age_preview)
print("\nGender columns:")
print(gender_preview)
# Choose appropriate columns for age and gender
age_cols = {'age_at_initial_pathologic_diagnosis', 'days_to_birth'}
gender_cols = {'gender'}

# Set age column - prefer direct age over days_to_birth if available
if 'age_at_initial_pathologic_diagnosis' in age_cols:
    age_col = 'age_at_initial_pathologic_diagnosis'
elif 'days_to_birth' in age_cols:
    age_col = 'days_to_birth'
else:
    age_col = None

# Set gender column
gender_col = 'gender' if 'gender' in gender_cols else None

# Print selected columns
print(f"Selected age column: {age_col}")
print(f"Selected gender column: {gender_col}")
# 1. Extract and standardize clinical features
# First create trait labels using sample IDs, then add demographics if available
clinical_features = tcga_select_clinical_features(
    clinical_df, 
    trait=trait,
    age_col='age_at_initial_pathologic_diagnosis',
    gender_col='gender'
)

# 2. Normalize gene symbols and save
normalized_gene_df = normalize_gene_symbols_in_index(genetic_df)
os.makedirs(os.path.dirname(out_gene_data_file), exist_ok=True)
normalized_gene_df.to_csv(out_gene_data_file)

# 3. Link clinical and genetic data
linked_data = pd.concat([clinical_features, normalized_gene_df.T], axis=1)

# 4. Handle missing values systematically
linked_data = handle_missing_values(linked_data, trait)

# 5. Check for bias in trait and demographic features
trait_biased, linked_data = judge_and_remove_biased_features(linked_data, trait)

# 6. Validate data quality and save cohort info
note = "Contains molecular data from tumor and normal samples with patient demographics."
is_usable = validate_and_save_cohort_info(
    is_final=True,
    cohort="TCGA",
    info_path=json_path,
    is_gene_available=True,
    is_trait_available=True,
    is_biased=trait_biased,
    df=linked_data,
    note=note
)

# 7. Save linked data if usable
if is_usable:
    os.makedirs(os.path.dirname(out_data_file), exist_ok=True)
    linked_data.to_csv(out_data_file)