p3/preprocess/Underweight/code/TCGA.py

# Path Configuration
from tools.preprocess import *

# Processing context
trait = "Underweight"

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

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

# Review all cohort directories
cohorts = os.listdir(tcga_root_dir)
cohorts = [c for c in cohorts if not c.startswith('.') and not c.endswith('.ipynb')]

# For underweight status, we can use cancer datasets since:
# 1. Cancer and its treatments often affect body weight 
# 2. Many cancer studies track patient weight/BMI
# 3. TCGA only contains cancer datasets

# Choose pancreatic cancer cohort because:
# 1. Weight loss and underweight are common in pancreatic cancer patients
# 2. Weight status is particularly relevant to prognosis and treatment
cohort_dir = "TCGA_Pancreatic_Cancer_(PAAD)"
cohort_path = os.path.join(tcga_root_dir, cohort_dir)

# Get clinical and genetic data file paths
clinical_file, genetic_file = tcga_get_relevant_filepaths(cohort_path)

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

# Print clinical data columns  
print("Clinical data columns:")
print(clinical_df.columns.tolist())
# 1. Define candidate columns for age and gender
candidate_age_cols = ['age_at_initial_pathologic_diagnosis', 'days_to_birth']
candidate_gender_cols = ['gender']

# Get file paths
cohort_dir = os.path.join(tcga_root_dir, "PAAD")
clinical_file_path, _ = tcga_get_relevant_filepaths(cohort_dir)

# 2. Preview first 5 values
clinical_data = pd.read_csv(clinical_file_path, index_col=0)

# Preview age columns
age_preview = {}
for col in candidate_age_cols:
    age_preview[col] = clinical_data[col].head().tolist()
print("Age columns preview:", age_preview)

# Preview gender columns
gender_preview = {}
for col in candidate_gender_cols:
    gender_preview[col] = clinical_data[col].head().tolist()
print("Gender columns preview:", gender_preview)
# Review all cohort directories
cohorts = os.listdir(tcga_root_dir)
cohorts = [c for c in cohorts if not c.startswith('.') and not c.endswith('.ipynb')]

# For underweight status, we can use cancer datasets since:
# 1. Cancer and its treatments often affect body weight 
# 2. Many cancer studies track patient weight/BMI
# 3. TCGA only contains cancer datasets

# Choose pancreatic cancer cohort because:
# 1. Weight loss and underweight are common in pancreatic cancer patients
# 2. Weight status is particularly relevant to prognosis and treatment
cohort_dir = "TCGA_Pancreatic_Cancer_(PAAD)"
cohort_path = os.path.join(tcga_root_dir, cohort_dir)

# Get clinical and genetic data file paths
clinical_file, genetic_file = tcga_get_relevant_filepaths(cohort_path)

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

# Print clinical data columns  
print("Clinical data columns:")
print(clinical_df.columns.tolist())
# Select age column (choose age_at_initial_pathologic_diagnosis as it directly represents patient age)
age_col = "age_at_initial_pathologic_diagnosis"

# Select gender column (gender is clearly named and appropriate for demographics)
gender_col = "gender"

# Print chosen columns
print(f"Selected age column: {age_col}")
print(f"Selected gender column: {gender_col}")
# Select appropriate demographic columns
age_col = 'age_at_initial_pathologic_diagnosis'  # This is more directly usable than days_to_birth
gender_col = 'gender'

# 1. Extract and standardize clinical features
selected_clinical_df = tcga_select_clinical_features(clinical_df, trait, age_col, gender_col)

# 2. Normalize gene symbols in genetic data
normalized_genetic_df = normalize_gene_symbols_in_index(genetic_df)
os.makedirs(os.path.dirname(out_gene_data_file), exist_ok=True)
normalized_genetic_df.to_csv(out_gene_data_file)

# 3. Link clinical and genetic data
linked_data = pd.merge(selected_clinical_df, normalized_genetic_df.T, left_index=True, right_index=True)

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

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

# 6. Validate and save cohort info 
note = f"Sample size after preprocessing: {len(linked_data)}. Number of genes: {len(linked_data.columns) - 3}"
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=is_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)
    print(f"Linked data saved to {out_data_file}")
    print("Shape of final linked data:", linked_data.shape)
else:
    print("Dataset was found to be unusable and was not saved")