output/preprocess/Asthma/code/GSE123086.py

# Path Configuration
from tools.preprocess import *

# Processing context
trait = "Asthma"
cohort = "GSE123086"

# Input paths
in_trait_dir = "../DATA/GEO/Asthma"
in_cohort_dir = "../DATA/GEO/Asthma/GSE123086"

# Output paths
out_data_file = "./output/z1/preprocess/Asthma/GSE123086.csv"
out_gene_data_file = "./output/z1/preprocess/Asthma/gene_data/GSE123086.csv"
out_clinical_data_file = "./output/z1/preprocess/Asthma/clinical_data/GSE123086.csv"
json_path = "./output/z1/preprocess/Asthma/cohort_info.json"


# Step 1: Initial Data Loading
from tools.preprocess import *
# 1. Identify the paths to the SOFT file and the matrix file
soft_file, matrix_file = geo_get_relevant_filepaths(in_cohort_dir)

# 2. Read the matrix file to obtain background information and sample characteristics data
background_prefixes = ['!Series_title', '!Series_summary', '!Series_overall_design']
clinical_prefixes = ['!Sample_geo_accession', '!Sample_characteristics_ch1']
background_info, clinical_data = get_background_and_clinical_data(matrix_file, background_prefixes, clinical_prefixes)

# 3. Obtain the sample characteristics dictionary from the clinical dataframe
sample_characteristics_dict = get_unique_values_by_row(clinical_data)

# 4. Explicitly print out all the background information and the sample characteristics dictionary
print("Background Information:")
print(background_info)
print("Sample Characteristics Dictionary:")
print(sample_characteristics_dict)

# Step 2: Dataset Analysis and Clinical Feature Extraction
import re

# 1) Gene expression data availability (Agilent microarray gene expression per background)
is_gene_available = True

# 2) Variable availability and conversion functions

# Decide rows based on the Sample Characteristics Dictionary in the prompt:
# - trait_row: primary diagnosis -> row 1
# - gender_row: contains 'Sex:' (row 2 has Sex plus some diagnosis2; handle non-sex values in converter)
# - age_row: rows 3/4 show ages; choose row 3 (handle non-age values in converter)
trait_row = 1
gender_row = 2
age_row = 3

# Conversion helpers
def _after_colon(x: str) -> str:
    if x is None:
        return ""
    parts = str(x).split(":", 1)
    return parts[1].strip() if len(parts) > 1 else str(x).strip()

def convert_trait(x):
    # Binary: 1 for trait present (Asthma), 0 for all others (including healthy controls and other diseases)
    v = _after_colon(x).strip().lower()
    if not v:
        return None
    # Match trait name robustly
    # We only consider "primary diagnosis" row, but keep a generic check
    if "asthma" in v:
        return 1
    # If it's clearly a known non-trait value (e.g., healthy control or other diseases), map to 0
    non_trait_keywords = [
        "healthy_control", "obesity", "seasonal_allergic_rhinitis", "psoriasis",
        "crohn", "influenza", "ulcerative_colitis", "atherosclerosis",
        "breast_cancer", "type_1_diabetes", "chronic_lymphocytic_leukemia",
        "atopic_eczema", "acute_tonsillitis"
    ]
    if any(k in v for k in non_trait_keywords):
        return 0
    return None

def convert_age(x):
    # Continuous: extract numeric age in years; invalid entries -> None
    v = _after_colon(x)
    # Some cells in row 3 may contain "Sex: ..." -> return None
    m = re.search(r"(-?\d+(?:\.\d+)?)", v)
    if not m:
        return None
    try:
        age_val = float(m.group(1))
        if 0 <= age_val <= 120:
            return age_val
        return None
    except Exception:
        return None

def convert_gender(x):
    # Binary: female->0, male->1; unknown -> None
    v = _after_colon(x).strip().lower()
    if v in ["female", "f"]:
        return 0
    if v in ["male", "m"]:
        return 1
    return None

# 3) Save metadata using initial filtering
is_trait_available = trait_row is not None
_ = validate_and_save_cohort_info(
    is_final=False,
    cohort=cohort,
    info_path=json_path,
    is_gene_available=is_gene_available,
    is_trait_available=is_trait_available
)

# 4) Clinical feature extraction (only if trait_row is available)
if trait_row is not None:
    selected_clinical_df = geo_select_clinical_features(
        clinical_df=clinical_data,
        trait=trait,
        trait_row=trait_row,
        convert_trait=convert_trait,
        age_row=age_row,
        convert_age=convert_age,
        gender_row=gender_row,
        convert_gender=convert_gender
    )
    preview = preview_df(selected_clinical_df, n=5)
    print(preview)
    # Save clinical features
    os.makedirs(os.path.dirname(out_clinical_data_file), exist_ok=True)
    selected_clinical_df.to_csv(out_clinical_data_file)

# Step 3: Gene Data Extraction
# 1. Use the get_genetic_data function from the library to get the gene_data from the matrix_file previously defined.
gene_data = get_genetic_data(matrix_file)

# 2. Print the first 20 row IDs (gene or probe identifiers) for future observation.
print(gene_data.index[:20])

# Step 4: Gene Identifier Review
print("requires_gene_mapping = True")

# Step 5: Gene Annotation
# 1. Use the 'get_gene_annotation' function from the library to get gene annotation data from the SOFT file.
gene_annotation = get_gene_annotation(soft_file)

# 2. Use the 'preview_df' function from the library to preview the data and print out the results.
print("Gene annotation preview:")
print(preview_df(gene_annotation))

# Step 6: Gene Identifier Mapping
import re
import pandas as pd

# Preserve original probe-/ID-level expression
expr_df = gene_data.copy()

def normalize_id_series(s: pd.Series) -> pd.Series:
    s = s.astype(str).str.strip()
    return s.str.replace(r'\.0$', '', regex=True)

# 1) Choose ID column from annotation that best matches expression IDs (should be "ID")
expr_ids = set(expr_df.index.astype(str).str.strip())
best_id_col = None
best_overlap = -1
for col in gene_annotation.columns:
    cand = normalize_id_series(gene_annotation[col])
    overlap = cand.isin(expr_ids).sum()
    if overlap > best_overlap:
        best_overlap = overlap
        best_id_col = col

# Prefer explicit 'ID' if reasonable
if 'ID' in gene_annotation.columns:
    cand = normalize_id_series(gene_annotation['ID'])
    overlap = cand.isin(expr_ids).sum()
    if overlap >= best_overlap * 0.95:
        best_id_col = 'ID'

# 2) Use ENTREZ_GENE_ID as the gene identifier since symbol columns are absent
if 'ENTREZ_GENE_ID' not in gene_annotation.columns:
    raise ValueError("ENTREZ_GENE_ID column not found in annotation; cannot proceed with Entrez mapping.")

print(f"Chosen ID column: {best_id_col}")
print("Chosen Gene column: ENTREZ_GENE_ID (Entrez IDs)")

# 3) Build a clean, 1:1 mapping from ID -> Entrez (numeric-only), avoiding token explosion
annotation_for_map = gene_annotation.loc[:, [best_id_col, 'ENTREZ_GENE_ID']].copy()
annotation_for_map[best_id_col] = normalize_id_series(annotation_for_map[best_id_col])

# Keep only IDs present in expression
annotation_for_map = annotation_for_map[annotation_for_map[best_id_col].isin(expr_df.index)]

# Extract pure numeric Entrez IDs; drop rows without a valid numeric Entrez
def extract_numeric_entrez(x):
    if pd.isna(x):
        return None
    m = re.search(r'\d+', str(x))
    return m.group(0) if m else None

annotation_for_map['Gene'] = annotation_for_map['ENTREZ_GENE_ID'].apply(extract_numeric_entrez)
annotation_for_map = annotation_for_map.dropna(subset=['Gene'])

# Reduce to necessary columns and remove duplicates
mapping_df = annotation_for_map.loc[:, [best_id_col, 'Gene']].rename(columns={best_id_col: 'ID'})
mapping_df['ID'] = mapping_df['ID'].astype(str).str.strip()
mapping_df['Gene'] = mapping_df['Gene'].astype(str).str.strip()
mapping_df = mapping_df.drop_duplicates(subset=['ID', 'Gene'])

# 4) Map expression to Entrez gene IDs (1:1 in this dataset) and aggregate
mapped = mapping_df.set_index('ID').join(expr_df, how='inner')
expr_cols = [c for c in mapped.columns if c != 'Gene']
gene_data = mapped.groupby('Gene')[expr_cols].sum()

# 5) Sanity check gene count to catch mapping explosions
n_genes = gene_data.shape[0]
n_samples = gene_data.shape[1]
print(f"Gene-level data shape: {gene_data.shape} (genes x samples)")
if n_genes < 5000 or n_genes > 80000:
    raise ValueError(f"Implausible gene count after mapping: {n_genes}. Check mapping logic and annotation parsing.")

# Step 7: Data Normalization and Linking
import os

# 1. Normalize gene data only if index appears to be gene symbols; otherwise keep Entrez IDs
idx = gene_data.index.astype(str)
digit_ratio = idx.str.fullmatch(r'\d+').mean()  # proportion of purely numeric IDs
note = ""

if digit_ratio < 0.5:
    # Likely gene symbols: normalize using synonym information
    normalized_gene_data = normalize_gene_symbols_in_index(gene_data)
    note = "INFO: Gene symbols detected; normalized using synonym dictionary."
else:
    # Likely Entrez IDs: skip normalization
    normalized_gene_data = gene_data.copy()
    note = "INFO: Gene matrix indexed by Entrez Gene IDs; gene symbol normalization skipped."

# Ensure output directory exists and save gene data
os.makedirs(os.path.dirname(out_gene_data_file), exist_ok=True)
normalized_gene_data.to_csv(out_gene_data_file)

# 2. Link the clinical and genetic data
linked_data = geo_link_clinical_genetic_data(selected_clinical_df, normalized_gene_data)

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

# 4. Determine bias and remove biased demographic features
is_trait_biased, unbiased_linked_data = judge_and_remove_biased_features(linked_data, trait)

# 5. Final validation and save cohort info
is_usable = validate_and_save_cohort_info(
    True, cohort, json_path, True, True, is_trait_biased, unbiased_linked_data, note=note
)

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