Molfeat 使用示例

# Molfeat 使用示例 本文档提供了常见 molfeat 用例的实践示例。 ## 安装 bash # 推荐：使用 conda/mamba mamba install -c conda-forge molfeat # 替代方案：使用 pip pip install molfeat # 安装所有可选依赖 pip install "molfeat[all]" # 安装特定依赖 pip install "molfeat[dgl]" # 用于 GNN 模型 pip install "molfeat[graphormer]" # 用于 Graphormer pip install "molfeat[transformer]" # 用于 ChemBERTa, ChemGPT --- ## 快速入门 ### 基本特征化工作流 python import datamol as dm from molfeat.calc import FPCalculator from molfeat.trans import MoleculeTransformer # 加载示例数据 data = dm.data.freesolv().sample(100).smiles.values # 单分子特征化 calc = FPCalculator("ecfp") features_single = calc(data[0]) print(f"单分子特征形状: {features_single.shape}") # 输出: (2048,) # 并行批处理特征化 transformer = MoleculeTransformer(calc, n_jobs=-1) features_batch = transformer(data) print(f"批处理特征形状: {features_batch.shape}") # 输出: (100, 2048) --- ## 计算器示例 ### 指纹计算器 python from molfeat.calc import FPCalculator # ECFP (扩展连接性指纹) ecfp = FPCalculator("ecfp", radius=3, fpSize=2048) fp = ecfp("CCO") # 乙醇 print(f"ECFP 形状: {fp.shape}") # (2048,) # MACCS 键 maccs = FPCalculator("maccs") fp = maccs("c1ccccc1") # 苯 print(f"MACCS 形状: {fp.shape}") # (167,) # 基于计数的指纹 ecfp_count = FPCalculator("ecfp-count", radius=3) fp_count = ecfp_count("CC(C)CC(C)C") # 非二进制计数 # MAP4 指纹 map4 = FPCalculator("map4") fp = map4("CC(=O)Oc1ccccc1C(=O)O") # 阿司匹林 ### 描述符计算器 python from molfeat.calc import RDKitDescriptors2D, MordredDescriptors # RDKit 2D 描述符 (200+ 属性) desc2d = RDKitDescriptors2D() descriptors = desc2d("CCO") print(f"2D 描述符数量: {len(descriptors)}") # 获取描述符名称 names = desc2d.columns print(f"前 5 个描述符: {names[:5]}") # Mordred 描述符 (1800+ 属性) mordred = MordredDescriptors() descriptors = mordred("c1ccccc1O") # 苯酚 print(f"Mordred 描述符: {len(descriptors)}") ### 药效团计算器 python from molfeat.calc import CATSCalculator # 2D CATS 描述符 cats = CATSCalculator(mode="2D", scale="raw") descriptors = cats("CC(C)Cc1ccc(C)cc1C") # 伞花烃 print(f"CATS "},