output/preprocess/Bladder_Cancer/code/TCGA.py

# Path Configuration
from tools.preprocess import *

# Processing context
trait = "Bladder_Cancer"

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

# Output paths
out_data_file = "./output/z1/preprocess/Bladder_Cancer/TCGA.csv"
out_gene_data_file = "./output/z1/preprocess/Bladder_Cancer/gene_data/TCGA.csv"
out_clinical_data_file = "./output/z1/preprocess/Bladder_Cancer/clinical_data/TCGA.csv"
json_path = "./output/z1/preprocess/Bladder_Cancer/cohort_info.json"


# Step 1: Initial Data Loading
import os

# Step 1: Select the most appropriate TCGA cohort directory for Bladder Cancer
subdirs = [d for d in os.listdir(tcga_root_dir) if os.path.isdir(os.path.join(tcga_root_dir, d))]
candidates = [d for d in subdirs if 'bladder' in d.lower()]

selected_dir = None
if candidates:
    # Prefer directory explicitly containing BLCA code
    blca_candidates = [d for d in candidates if 'blca' in d.lower()]
    selected_dir = blca_candidates[0] if blca_candidates else candidates[0]

if selected_dir is None:
    # No suitable directory found; record and exit early
    validate_and_save_cohort_info(
        is_final=False,
        cohort="TCGA",
        info_path=json_path,
        is_gene_available=False,
        is_trait_available=False
    )
else:
    tcga_cohort_dir = os.path.join(tcga_root_dir, selected_dir)

    # Step 2: Identify clinical and genetic data file paths
    clinical_file_path, genetic_file_path = tcga_get_relevant_filepaths(tcga_cohort_dir)

    # Step 3: Load both files as DataFrames
    clinical_df = pd.read_csv(clinical_file_path, sep='\t', index_col=0, low_memory=False)
    genetic_df = pd.read_csv(genetic_file_path, sep='\t', index_col=0, low_memory=False)

    # Step 4: Print the column names of the clinical data
    print(list(clinical_df.columns))

# Step 2: Find Candidate Demographic Features
import re

# Identify candidate columns for age and gender from available clinical data columns
if 'clinical_df' in globals():
    available_cols = list(clinical_df.columns)
else:
    available_cols = []

lower_cols = {c: c.lower() for c in available_cols}

age_pattern = re.compile(r'(^|_)age($|_)')
gender_pattern = re.compile(r'(^|_)(gender|sex)($|_)')

candidate_age_cols = []
for c in available_cols:
    lc = lower_cols[c]
    if 'stage' in lc:
        continue  # exclude false positives like 'pathologic_stage'
    if age_pattern.search(lc) or ('birth' in lc):
        candidate_age_cols.append(c)

candidate_gender_cols = [c for c in available_cols if gender_pattern.search(lower_cols[c])]

# Strictly required output format for candidate lists
print(f"candidate_age_cols = {candidate_age_cols}")
print(f"candidate_gender_cols = {candidate_gender_cols}")

# Preview extracted candidate columns if available
if 'clinical_df' in globals() and candidate_age_cols:
    age_preview = preview_df(clinical_df[candidate_age_cols], n=5)
    print(age_preview)

if 'clinical_df' in globals() and candidate_gender_cols:
    gender_preview = preview_df(clinical_df[candidate_gender_cols], n=5)
    print(gender_preview)

# Step 3: Select Demographic Features
# Select age and gender columns based on candidate lists and previewed values

# Fallback previews from the previous step (only set if not already available)
if 'age_values_dict' not in globals():
    age_values_dict = {
        'age_at_initial_pathologic_diagnosis': [63, 66, 69, 59, 83],
        'age_began_smoking_in_years': [20.0, 15.0, float('nan'), float('nan'), 30.0],
        'days_to_birth': [-23323.0, -24428.0, -25259.0, -21848.0, -30520.0],
    }
if 'gender_values_dict' not in globals():
    gender_values_dict = {
        'gender': ['MALE', 'MALE', 'MALE', 'FEMALE', 'MALE']
    }

# Ensure candidate lists are available (use previous step's output if missing)
if 'candidate_age_cols' not in globals():
    candidate_age_cols = ['age_at_initial_pathologic_diagnosis', 'age_began_smoking_in_years', 'days_to_birth']
if 'candidate_gender_cols' not in globals():
    candidate_gender_cols = ['gender']

# Selection logic
age_col = None
gender_col = None

# Prefer direct age at diagnosis over smoking age or days_to_birth
preferred_age_order = [
    'age_at_initial_pathologic_diagnosis',
    'age_at_diagnosis',
    'age'
]
for c in preferred_age_order:
    if c in candidate_age_cols:
        age_col = c
        break
# If none of the preferred are present but days_to_birth exists, we could process it later; for now, we still prefer None over unsuitable columns
if age_col is None and 'days_to_birth' in candidate_age_cols:
    # Keep None here since days_to_birth requires transformation; we will only use if absolutely necessary
    pass

# Prefer 'gender' or 'sex'
preferred_gender_order = ['gender', 'sex']
for c in preferred_gender_order:
    if c in candidate_gender_cols:
        gender_col = c
        break

# Validate against clinical_df if available
if 'clinical_df' in globals():
    if age_col is not None and age_col not in clinical_df.columns:
        age_col = None
    if gender_col is not None and gender_col not in clinical_df.columns:
        gender_col = None

# Prepare previews for printing
def preview_values(col_name, values_dict):
    if 'clinical_df' in globals() and col_name is not None and col_name in clinical_df.columns:
        return clinical_df[col_name].head(5).tolist()
    return values_dict.get(col_name, None) if col_name is not None else None

age_preview = preview_values(age_col, age_values_dict)
gender_preview = preview_values(gender_col, gender_values_dict)

# Explicitly print out chosen columns and their previews
print("Selected age_col:", age_col)
print("age_col first 5 values:", age_preview)
print("Selected gender_col:", gender_col)
print("gender_col first 5 values:", gender_preview)

# Step 4: Feature Engineering and Validation
import os
import json

# Helper: sanitize objects to be JSON-serializable
def _to_builtin(obj):
    # Fast path for already serializable basic types
    if obj is None or isinstance(obj, (bool, int, float, str)):
        return obj
    # Containers
    if isinstance(obj, dict):
        return {str(_to_builtin(k)): _to_builtin(v) for k, v in obj.items()}
    if isinstance(obj, (list, tuple, set)):
        return [_to_builtin(v) for v in obj]
    # Pandas / numpy scalars
    try:
        # numpy/pandas scalars often have .item()
        if hasattr(obj, 'item'):
            return _to_builtin(obj.item())
    except Exception:
        pass
    # Pandas timestamps/timedeltas or other objects: convert to string
    try:
        return json.loads(json.dumps(obj))
    except Exception:
        return str(obj)

def validate_and_save_cohort_info_safe(is_final: bool, cohort: str, info_path: str, is_gene_available: bool,
                                       is_trait_available: bool, is_biased: Optional[bool] = None,
                                       df: Optional[pd.DataFrame] = None, note: str = '') -> bool:
    """
    Wrapper around validate_and_save_cohort_info that sanitizes values to avoid JSON serialization errors.
    Mirrors the original function's logic.
    """
    is_usable = False
    if not is_final:
        if is_gene_available and is_trait_available:
            return is_usable
        else:
            new_record = {"is_usable": False,
                          "is_gene_available": is_gene_available,
                          "is_trait_available": is_trait_available,
                          "is_available": False,
                          "is_biased": None,
                          "has_age": None,
                          "has_gender": None,
                          "sample_size": None,
                          "note": None}
    else:
        if (df is None) or (is_biased is None):
            raise ValueError("For final data validation, 'df' and 'is_biased' must be provided.")
        if len(df) <= 0 or len(df.columns) <= 4:
            print(f"Abnormality detected in the cohort: {cohort}. Preprocessing failed.")
            is_gene_available = False
            if len(df) <= 0:
                is_trait_available = False
        is_available = bool(is_gene_available and is_trait_available)
        is_usable = bool(is_available and (is_biased is False))
        new_record = {"is_usable": is_usable,
                      "is_gene_available": bool(is_gene_available),
                      "is_trait_available": bool(is_trait_available),
                      "is_available": is_available,
                      "is_biased": (bool(is_biased) if is_available else None),
                      "has_age": ("Age" in df.columns if is_available else None),
                      "has_gender": ("Gender" in df.columns if is_available else None),
                      "sample_size": (int(len(df)) if is_available else None),
                      "note": note}

    trait_directory = os.path.dirname(info_path)
    os.makedirs(trait_directory, exist_ok=True)
    if not os.path.exists(info_path):
        with open(info_path, 'w') as file:
            json.dump({}, file)
        print(f"A new JSON file was created at: {info_path}")

    with open(info_path, "r") as file:
        try:
            records = json.load(file)
        except json.JSONDecodeError:
            records = {}
    records[cohort] = new_record

    # Sanitize before dump
    records_sanitized = _to_builtin(records)

    temp_path = info_path + ".tmp"
    try:
        with open(temp_path, 'w') as file:
            json.dump(records_sanitized, file)
        os.replace(temp_path, info_path)
    except Exception as e:
        print(f"An error occurred: {e}")
        if os.path.exists(temp_path):
            os.remove(temp_path)
        raise

    return is_usable


# 1) Extract and standardize clinical features (trait, Age, Gender)
age_col = age_col if 'age_col' in globals() else None
gender_col = gender_col if 'gender_col' in globals() else None

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

# Save standardized clinical data for inspection
os.makedirs(os.path.dirname(out_clinical_data_file), exist_ok=True)
selected_clinical_df.to_csv(out_clinical_data_file)

# 2) Normalize gene symbols and save gene data (subset to BLCA samples present in clinical data)
blca_samples = selected_clinical_df.index.intersection(genetic_df.columns)
genetic_sub = genetic_df.loc[:, blca_samples]

gene_norm = normalize_gene_symbols_in_index(genetic_sub)

os.makedirs(os.path.dirname(out_gene_data_file), exist_ok=True)
gene_norm.to_csv(out_gene_data_file)

# 3) Link clinical and genetic data on sample IDs
linked_data = selected_clinical_df.join(gene_norm.T, how='inner')

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

# 5) Determine bias in trait and demographics; drop biased demographics
trait_biased, processed_df = judge_and_remove_biased_features(processed_df, trait)

# 6) Final quality validation and save cohort info (use safe wrapper to avoid JSON serialization issues)
covariate_cols = [trait, 'Age', 'Gender']
gene_cols_in_processed = [c for c in processed_df.columns if c not in covariate_cols]
is_gene_available = len(gene_cols_in_processed) > 0
is_trait_available = (trait in processed_df.columns) and processed_df[trait].notna().any()

note = "INFO: Cohort TCGA_BLCA; trait=Tumor(1) vs Normal(0) by TCGA sample type (01-09 vs 10-19). Gene symbols normalized via NCBI synonyms; duplicates averaged."

is_usable = validate_and_save_cohort_info_safe(
    is_final=True,
    cohort="TCGA",
    info_path=json_path,
    is_gene_available=is_gene_available,
    is_trait_available=is_trait_available,
    is_biased=trait_biased,
    df=processed_df,
    note=note
)

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