""" Exporta muestras (head 10) de los datos HDF5 a archivos de texto. """ import h5py import numpy as np from pathlib import Path DATA_DIR = Path("data") OUTPUT_DIR = Path("encoded_data_format") OUTPUT_DIR.mkdir(exist_ok=True) def decode_if_bytes(arr): """Decodifica arrays de bytes a strings.""" if arr.dtype.kind == 'S': # byte string return [x.decode('utf-8') for x in arr] return list(arr) def export_reference_data(): """Exporta head 10 de los datos de referencia.""" ref_path = DATA_DIR / "ref.h5" output_path = OUTPUT_DIR / "reference-data-format.txt" with open(output_path, 'w', encoding='utf-8') as f: f.write("=" * 80 + "\n") f.write("DATOS DE REFERENCIA (ref.h5)\n") f.write("Contiene perfiles de expresion por tipo celular\n") f.write("=" * 80 + "\n\n") with h5py.File(ref_path, 'r') as h5: for ref_type in h5.keys(): f.write(f"\n{'='*60}\n") f.write(f"[{ref_type}]\n") f.write(f"{'='*60}\n") group = h5[ref_type] # Mostrar keys disponibles f.write(f"Keys disponibles: {list(group.keys())}\n\n") # Datos principales if 'data' in group: data = group['data'][:] f.write(f"Shape de 'data': {data.shape}\n") f.write(f"dtype: {data.dtype}\n\n") # Head 10 filas x 10 columnas f.write("Head 10 filas x 10 columnas:\n") head = data[:10, :10] if data.ndim == 2 else data[:10] f.write(np.array2string(head, precision=4, suppress_small=True)) f.write("\n\n") # Nombres de genes (filas) if 'genes' in group: genes = decode_if_bytes(group['genes'][:]) f.write(f"Genes (primeros 10): {genes[:10]}\n") f.write(f"Total genes: {len(genes)}\n\n") # Nombres de tipos celulares o muestras (columnas) for col_key in ['cell_types', 'celltypes', 'samples', 'colnames']: if col_key in group: cols = decode_if_bytes(group[col_key][:]) f.write(f"{col_key}: {cols[:10]}\n") f.write(f"Total {col_key}: {len(cols)}\n\n") print(f"[OK] Guardado: {output_path}") def export_ground_truth(): """Exporta head 10 de los ground truth.""" output_path = OUTPUT_DIR / "ground-truth.txt" gt_files = list(DATA_DIR.glob("groundtruth*.h5")) with open(output_path, 'w', encoding='utf-8') as f: f.write("=" * 80 + "\n") f.write("GROUND TRUTH (Proporciones reales)\n") f.write("Cada columna es un tipo celular, cada fila una muestra\n") f.write("Los valores son proporciones (deben sumar 1 por fila)\n") f.write("=" * 80 + "\n\n") for gt_path in sorted(gt_files): f.write(f"\n{'='*60}\n") f.write(f"[{gt_path.name}]\n") f.write(f"{'='*60}\n") with h5py.File(gt_path, 'r') as h5: # El ground truth suele estar en /groundtruth/ if 'groundtruth' in h5: group = h5['groundtruth'] else: group = h5 f.write(f"Keys: {list(group.keys())}\n\n") if 'data' in group: data = group['data'][:] f.write(f"Shape: {data.shape} (muestras x tipos_celulares)\n") f.write(f"dtype: {data.dtype}\n\n") # Head 10 f.write("Head 10 filas:\n") head = data[:10, :] f.write(np.array2string(head, precision=4, suppress_small=True)) f.write("\n\n") # Verificar que suman 1 row_sums = data.sum(axis=1) f.write(f"Suma por fila (primeras 10): {row_sums[:10]}\n\n") # Nombres de tipos celulares for key in ['genes', 'cell_types', 'celltypes']: if key in group: names = decode_if_bytes(group[key][:]) f.write(f"{key}: {names}\n") # Nombres de muestras if 'samples' in group: samples = decode_if_bytes(group['samples'][:]) f.write(f"samples (primeras 10): {samples[:10]}\n") print(f"[OK] Guardado: {output_path}") def export_mixes(): """Exporta head 10 de los datos de mezclas a deconvolucionar.""" output_path = OUTPUT_DIR / "convulsioned.txt" mix_files = list(DATA_DIR.glob("mixes*.h5")) with open(output_path, 'w', encoding='utf-8') as f: f.write("=" * 80 + "\n") f.write("DATOS A DECONVOLUCIONAR (mixes)\n") f.write("Cada columna es una muestra bulk, cada fila un gen\n") f.write("Tu modelo debe estimar las proporciones de tipos celulares\n") f.write("=" * 80 + "\n\n") for mix_path in sorted(mix_files): f.write(f"\n{'='*60}\n") f.write(f"[{mix_path.name}]\n") f.write(f"{'='*60}\n") with h5py.File(mix_path, 'r') as h5: f.write(f"Keys principales: {list(h5.keys())}\n\n") # Explorar cada modalidad (mix_rna, mix_met, etc.) for modal_key in h5.keys(): group = h5[modal_key] f.write(f"--- {modal_key} ---\n") if isinstance(group, h5py.Group): f.write(f" Sub-keys: {list(group.keys())}\n") if 'data' in group: data = group['data'][:] f.write(f" Shape: {data.shape} (genes x muestras)\n") f.write(f" dtype: {data.dtype}\n") # Head 10x5 f.write(" Head 10 filas x 5 columnas:\n") if data.ndim == 2 and np.issubdtype(data.dtype, np.number): head = data[:10, :5] for row in head: f.write(" " + " ".join([f"{x:10.4f}" for x in row]) + "\n") elif data.ndim == 1 and np.issubdtype(data.dtype, np.number): f.write(" " + " ".join([f"{x:10.4f}" for x in data[:10]]) + "\n") elif data.ndim >= 1: # Es un array de strings u otro tipo items = decode_if_bytes(data[:10]) if data.dtype.kind == 'S' else list(data[:10]) f.write(f" {items}\n") else: f.write(f" Valor: {data}\n") f.write("\n") # Genes if 'genes' in group: genes = decode_if_bytes(group['genes'][:]) f.write(f" genes (primeros 10): {genes[:10]}\n") f.write(f" Total genes: {len(genes)}\n") # Samples if 'samples' in group: samples = decode_if_bytes(group['samples'][:]) f.write(f" samples: {samples[:10]}\n") f.write("\n") elif isinstance(group, h5py.Dataset): f.write(f" Shape: {group.shape}\n") f.write(f" dtype: {group.dtype}\n\n") print(f"[OK] Guardado: {output_path}") def main(): print("Exportando muestras de datos...") print("-" * 40) export_reference_data() export_ground_truth() export_mixes() print("-" * 40) print("Completado!") if __name__ == "__main__": main()