p3/preprocess/LDL_Cholesterol_Levels/code/TCGA.py

# Path Configuration
from tools.preprocess import *

# Processing context
trait = "LDL_Cholesterol_Levels"

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

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

# Cannot proceed with column identification without first having access to 
# the column names from the previous step's output

# For now, define empty candidates
candidate_age_cols = []  
candidate_gender_cols = []

preview_dict = {}
preview_dict
# 1. From the subdirectories list, none contain terms directly related to LDL cholesterol or lipid levels
# Therefore, we need to examine a proxy tissue/condition most related to cholesterol metabolism
# The liver is the primary organ for cholesterol metabolism, so we'll use liver cancer data
cohort_dir = os.path.join(tcga_root_dir, 'TCGA_Liver_Cancer_(LIHC)')

# 2. Get the clinical and genetic data 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())
# Define candidate columns
candidate_age_cols = ['age_at_initial_pathologic_diagnosis', 'days_to_birth', 'year_of_initial_pathologic_diagnosis']
candidate_gender_cols = ['gender']

# Use LIHC (Liver Cancer) data
cohort_dir = os.path.join(tcga_root_dir, "TCGA_Liver_Cancer_(LIHC)")

# Get clinical data path
clinical_file_path, _ = tcga_get_relevant_filepaths(cohort_dir)
clinical_df = pd.read_csv(clinical_file_path, index_col=0)

# Preview age columns
age_preview = {}
for col in candidate_age_cols:
    if col in clinical_df.columns:
        age_preview[col] = clinical_df[col].head(5).tolist()
print("Age columns preview:", age_preview)

# Preview gender columns 
gender_preview = {}
for col in candidate_gender_cols:
    if col in clinical_df.columns:
        gender_preview[col] = clinical_df[col].head(5).tolist()
print("\nGender columns preview:", gender_preview)
# Information from previous step
# Dictionaries containing sample values from candidate columns
age_candidates = {'age_at_initial_pathologic_diagnosis': [63, 53, 69, 65, 59], 'age_began_smoking_in_years': ['[Not Applicable]', '[Not Available]', '[Not Available]', '[Not Available]', '[Not Applicable]']}
gender_candidates = {'gender': ['FEMALE', 'FEMALE', 'FEMALE', 'MALE', 'MALE']}

# Select age column - choose 'age_at_initial_pathologic_diagnosis' as it has valid numeric values
age_col = 'age_at_initial_pathologic_diagnosis' if 'age_at_initial_pathologic_diagnosis' in age_candidates and all(isinstance(x, (int, float)) for x in age_candidates['age_at_initial_pathologic_diagnosis']) else None

# Select gender column - choose 'gender' if it contains valid gender values
gender_col = 'gender' if 'gender' in gender_candidates and all(isinstance(x, str) and x.upper() in ['MALE', 'FEMALE'] for x in gender_candidates['gender']) else None

# Print chosen columns
print(f"Selected age column: {age_col}")
print(f"Selected gender column: {gender_col}")
# 1. Extract and standardize clinical features
# First reload data with correct separator
clinical_file_path, genetic_file_path = tcga_get_relevant_filepaths(cohort_dir)
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')

# Use days_to_birth as a source for age calculation since LDL is a continuous trait
age_values = (-clinical_df['days_to_birth']/365).round()
age_values = age_values.fillna(age_values.mean()).astype(int)
clinical_df['age_at_initial_pathologic_diagnosis'] = age_values

selected_clinical_df = tcga_select_clinical_features(clinical_df, trait, age_col=age_col, gender_col=gender_col)

# 2. Normalize gene symbols
normalized_gene_df = normalize_gene_symbols_in_index(genetic_df)

# Save normalized gene data
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([selected_clinical_df, normalized_gene_df.T], axis=1)

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

# 5. Check for biased features and remove biased demographic features
is_trait_biased, cleaned_data = judge_and_remove_biased_features(linked_data, trait)

# 6. Validate data quality and save cohort info
note = "Data from TCGA Liver Cancer cohort used as proxy for LDL cholesterol studies due to liver's role in cholesterol metabolism. Age was calculated from days_to_birth for more accurate values."
is_usable = validate_and_save_cohort_info(
    is_final=True,
    cohort="TCGA_LIHC",
    info_path=json_path,
    is_gene_available=True,
    is_trait_available=True,
    is_biased=is_trait_biased,
    df=cleaned_data,
    note=note
)

# 7. Save linked data if usable
if is_usable:
    os.makedirs(os.path.dirname(out_data_file), exist_ok=True)
    cleaned_data.to_csv(out_data_file)
    print(f"Data saved to {out_data_file}")
else:
    print("Data quality validation failed. Dataset not saved.")
# 1. Extract and standardize clinical features
# First reload data with correct separator
clinical_file_path, genetic_file_path = tcga_get_relevant_filepaths(cohort_dir)
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')

# Define demographic columns based on inspection from previous steps
age_col = 'age_at_initial_pathologic_diagnosis'
gender_col = 'gender'

# Calculate age from days_to_birth for more accuracy
age_values = (-clinical_df['days_to_birth']/365).round()
age_values = age_values.fillna(age_values.mean()).astype(int)
clinical_df[age_col] = age_values

selected_clinical_df = tcga_select_clinical_features(clinical_df, trait, age_col=age_col, gender_col=gender_col)

# 2. Normalize gene symbols
normalized_gene_df = normalize_gene_symbols_in_index(genetic_df)

# Save normalized gene data
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([selected_clinical_df, normalized_gene_df.T], axis=1)

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

# 5. Check for biased features and remove biased demographic features
is_trait_biased, cleaned_data = judge_and_remove_biased_features(linked_data, trait)

# 6. Validate data quality and save cohort info
note = "Data from TCGA Liver Cancer cohort used as proxy for LDL cholesterol studies due to liver's role in cholesterol metabolism. Age was calculated from days_to_birth for more accurate values."
is_usable = validate_and_save_cohort_info(
    is_final=True,
    cohort="TCGA_LIHC",
    info_path=json_path,
    is_gene_available=True,
    is_trait_available=True,
    is_biased=is_trait_biased,
    df=cleaned_data,
    note=note
)

# 7. Save linked data if usable
if is_usable:
    os.makedirs(os.path.dirname(out_data_file), exist_ok=True)
    cleaned_data.to_csv(out_data_file)
    print(f"Data saved to {out_data_file}")
else:
    print("Data quality validation failed. Dataset not saved.")