#!/usr/bin/env python # -*- coding: utf-8 -*- """ 为未被现有 module 覆盖的基因创建额外的 module 确保所有模型基因都能参与训练 使用方法: python create_uncovered_modules.py \ --merged_modules module_data_all/merged_modules.json \ --output module_data_all/merged_modules_full.json \ --module_size 50 """ import json import os import csv import argparse def create_uncovered_modules( merged_modules_path, gene_list_path, gene_mapping_path, output_path, module_size=50, verbose=True ): """ 为未被覆盖的基因创建额外的 module Args: merged_modules_path: 已合并的 module 文件路径 gene_list_path: 模型基因列表路径 gene_mapping_path: 基因映射文件路径 output_path: 输出文件路径 module_size: 新 module 的大小 verbose: 是否打印详细信息 """ # 加载模型基因列表 with open(gene_list_path, 'r') as f: model_genes = set(line.strip() for line in f) # 加载已有的 modules with open(merged_modules_path, 'r') as f: modules = json.load(f) # 加载基因映射 (symbol -> ensembl) symbol_to_ens = {} ens_to_symbol = {} with open(gene_mapping_path, 'r') as f: reader = csv.DictReader(f) for row in reader: symbol = row.get('Approved symbol', '').strip() ens_id = row.get('Ensembl gene ID', '').strip() if symbol and ens_id: symbol_to_ens[symbol] = ens_id ens_to_symbol[ens_id] = symbol # 找出已被覆盖的基因 covered_genes_ens = set() for module_name, genes in modules.items(): for g in genes: ens_id = symbol_to_ens.get(g) if ens_id and ens_id in model_genes: covered_genes_ens.add(ens_id) elif g in model_genes: covered_genes_ens.add(g) # 找出未被覆盖的基因 uncovered_genes = model_genes - covered_genes_ens uncovered_genes = sorted(list(uncovered_genes)) if verbose: print(f"模型基因总数: {len(model_genes)}") print(f"已被覆盖的基因: {len(covered_genes_ens)} ({len(covered_genes_ens)/len(model_genes)*100:.1f}%)") print(f"未被覆盖的基因: {len(uncovered_genes)} ({len(uncovered_genes)/len(model_genes)*100:.1f}%)") # 为未覆盖的基因创建新的 modules # 使用 Ensembl ID(因为模型使用的是 Ensembl ID) new_modules = {} num_new_modules = (len(uncovered_genes) + module_size - 1) // module_size for i in range(num_new_modules): start_idx = i * module_size end_idx = min((i + 1) * module_size, len(uncovered_genes)) module_genes = uncovered_genes[start_idx:end_idx] # 尝试转换为 symbol(如果有映射的话),否则保持 Ensembl ID module_genes_final = [] for g in module_genes: symbol = ens_to_symbol.get(g) if symbol: module_genes_final.append(symbol) else: module_genes_final.append(g) new_modules[f"uncovered_module_{i+1:04d}"] = module_genes_final if verbose: print(f"\n创建了 {len(new_modules)} 个新 module (每个约 {module_size} 个基因)") # 合并原有 modules 和新 modules merged_modules = {**modules, **new_modules} # 保存 os.makedirs(os.path.dirname(output_path) if os.path.dirname(output_path) else '.', exist_ok=True) with open(output_path, 'w', encoding='utf-8') as f: json.dump(merged_modules, f, indent=2) if verbose: print(f"\n最终结果:") print(f" 原有 modules: {len(modules)}") print(f" 新增 modules: {len(new_modules)}") print(f" 总 modules: {len(merged_modules)}") print(f" 保存到: {output_path}") return merged_modules def main(): parser = argparse.ArgumentParser(description='Create modules for uncovered genes') parser.add_argument('--merged_modules', type=str, default='module_data_all/merged_modules.json', help='Path to merged modules JSON file') parser.add_argument('--gene_list', type=str, default='../selected_genes_27k.txt', help='Path to model gene list file') parser.add_argument('--gene_mapping', type=str, default='../gene2ens.csv', help='Path to gene mapping CSV file') parser.add_argument('--output', '-o', type=str, default='module_data_all/merged_modules_full.json', help='Output file path') parser.add_argument('--module_size', type=int, default=50, help='Size of new modules for uncovered genes') args = parser.parse_args() create_uncovered_modules( merged_modules_path=args.merged_modules, gene_list_path=args.gene_list, gene_mapping_path=args.gene_mapping, output_path=args.output, module_size=args.module_size ) if __name__ == '__main__': main()