p3/preprocess/Depression/code/TCGA.py

# Path Configuration
from tools.preprocess import *

# Processing context
trait = "Depression"

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

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

depression_related_terms = ["depression", "mental", "psychiatric", "psychological", "mood"]

# Check if any cohort contains depression-related data
found_relevant_data = False

for cohort in cohorts:
    cohort_dir = os.path.join(tcga_root_dir, cohort)
    
    try:
        clinical_file, genetic_file = tcga_get_relevant_filepaths(cohort_dir)
        clinical_df = pd.read_csv(clinical_file, sep='\t', index_col=0)
        
        # Check column names for relevant terms
        relevant_cols = [col for col in clinical_df.columns 
                        if any(term in col.lower() for term in depression_related_terms)]
                        
        if relevant_cols:
            # Check if these columns actually contain meaningful data
            non_null_counts = clinical_df[relevant_cols].count()
            if (non_null_counts > 0).any():
                print(f"\nFound depression-related data in {cohort}:")
                print(f"Relevant columns: {relevant_cols}")
                found_relevant_data = True
                break
            
    except:
        continue

# Record result in metadata
validate_and_save_cohort_info(
    is_final=False,
    cohort="TCGA",
    info_path=json_path,
    is_gene_available=True,  # TCGA always has gene expression data
    is_trait_available=found_relevant_data
)
# Define depression-related search terms
depression_related_terms = ["depression", "mental", "psychiatric", "psychological", "mood", "affect"]

# Initialize found_trait_data flag
found_trait_data = False 

# Get list of TCGA cohorts
cohorts = [d for d in os.listdir(tcga_root_dir) if os.path.isdir(os.path.join(tcga_root_dir, d))]

# Search through cohorts for depression-related data
for cohort in cohorts:
    cohort_dir = os.path.join(tcga_root_dir, cohort)
    
    try:
        # Get clinical and genetic file paths
        clinical_file, genetic_file = tcga_get_relevant_filepaths(cohort_dir)
        
        # Load clinical data and check column names
        clinical_df = pd.read_csv(clinical_file, sep='\t', index_col=0)
        genetic_df = pd.read_csv(genetic_file, sep='\t', index_col=0)
        
        # Look for depression-related columns
        relevant_cols = [col for col in clinical_df.columns 
                        if any(term in col.lower() for term in depression_related_terms)]
        
        if relevant_cols:
            # Check if columns contain non-null data
            non_null_counts = clinical_df[relevant_cols].count()
            if (non_null_counts > 0).any():
                print(f"\nFound depression-related data in {cohort}:")
                print(f"Relevant columns: {relevant_cols}")
                print("\nClinical data columns:")
                print(clinical_df.columns.tolist())
                found_trait_data = True
                break
                
    except Exception as e:
        continue

# Record results in metadata
validate_and_save_cohort_info(
    is_final=False,
    cohort="TCGA",
    info_path=json_path,
    is_gene_available=True,  # TCGA always has gene expression data
    is_trait_available=found_trait_data
)
# Define candidate columns based on column names containing age/gender related keywords
candidate_age_cols = ['age_at_initial_pathologic_diagnosis', 'days_to_birth', 'first_diagnosis_age_asth_ecz_hay_fev_mold_dust', 'first_diagnosis_age_of_animal_insect_allergy', 'first_diagnosis_age_of_food_allergy']
candidate_gender_cols = ['gender']

# Get clinical file path
clinical_file_path, _ = tcga_get_relevant_filepaths(os.path.join(tcga_root_dir, 'TCGA_lower_grade_glioma_and_glioblastoma_(GBMLGG)'))

# Read clinical data
clinical_df = pd.read_csv(clinical_file_path, sep='\t', index_col=0)

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

# Preview gender columns  
gender_preview = {}
for col in candidate_gender_cols:
    gender_preview[col] = clinical_df[col].head().tolist()
print("\nGender columns preview:")
print(gender_preview)
# Select age column by inspecting preview data
# 'age_at_initial_pathologic_diagnosis' has valid age values
age_col = 'age_at_initial_pathologic_diagnosis'

# Select gender column by inspecting preview data
# 'gender' contains valid gender values
gender_col = 'gender'

# Print chosen columns
print(f"Selected age column: {age_col}")
print(f"Selected gender column: {gender_col}")
# Define demographic columns discovered in previous steps
age_col = 'age_at_initial_pathologic_diagnosis'
gender_col = 'gender'

# Set up cohort directory and load data
cohort_dir = os.path.join(tcga_root_dir, 'TCGA_lower_grade_glioma_and_glioblastoma_(GBMLGG)')
clinical_file, genetic_file = tcga_get_relevant_filepaths(cohort_dir)
clinical_df = pd.read_csv(clinical_file, sep='\t', index_col=0)
genetic_df = pd.read_csv(genetic_file, sep='\t', index_col=0)

# Extract clinical features using mental_status_changes as depression indicator
clinical_features = tcga_select_clinical_features(
    clinical_df, 
    trait='mental_status_changes',  # Use mental status changes as depression indicator
    age_col=age_col,
    gender_col=gender_col
)

# Save processed clinical data
os.makedirs(os.path.dirname(out_clinical_data_file), exist_ok=True)
clinical_features.to_csv(out_clinical_data_file)

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

# Link clinical and genetic data
linked_data = pd.merge(
    clinical_features,
    normalized_gene_data.T,
    left_index=True,
    right_index=True,
    how='inner'
)

# Handle missing values
linked_data = handle_missing_values(linked_data, 'mental_status_changes')

# Check if trait or demographic features are biased and remove biased demographics
is_trait_biased, linked_data = judge_and_remove_biased_features(linked_data, 'mental_status_changes')

# Final validation and save metadata
notes = "Using TCGA glioma/glioblastoma (GBMLGG) data. Mental status changes used as depression indicator."
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_trait_biased,
    df=linked_data,
    note=notes
)

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