Supporting data for “Therapeutic Effect of Teng Qi Formula on Triple-negative Breast Cancer: Efficacy, Active principles, and Mechanisms”

In silico drug toxicity and interaction prediction workflow construction. The active compounds of herbal medicine defined here, are the natural products documented in TSCSP database for a certain herbal medicine, screened out based on the criteria (OB ≥ 0.3 and DL ≥ 0.18). These data were used to make a toxicity and drug interaction prediction for plant complexes. From the PubChem database, the mining of properties of active compounds was conducted firstly through a script coded in Python 3 (version 3.8.10) called “compound_properties_mining.py” using pubchempy and pandas packages. This script iterates over the “active_comp_pool_tcmsp.csv” dataset, specifically, the “Molecule Name” column, while fetching one “Molecule Name” at a time. The gathered property data of active compounds were written to a CSV file named “active_comp_proper_pubchem.csv”. Screened from the “Toxi_infor_sum.csv” file, the drug interaction information retrieved was separated and split into one “interaction” retrieve per row using a script named “drug_interactions_split.py” for further manual interpretation. The split data was stored in the file named “drug_interaction_pred_0.6171.csv”. The mining of similar compounds of active compounds was done through the web scraper script called “similar_comp_crawler.py”. This script iterated the “Active_compound_name” column and the “isomeric_smiles” column of the dataset storing the properties of active compounds. The isomeric SMILES 
code is posted as a query to the SwissSimilarity website (updated version issued in Dec. 2021), selecting“ Bioactive” compound class, choosing “ChEMBL (actives only)” natural product library, based on combined methods. All the data of similar compounds were stored in the file named “similar_comp_pool_swiss.csv”. Before toxicity and drug interaction information mining, using a script called “similar_compound_properties_mining.py”, the properties of similar compounds were collected with a similar method as the mining of properties of active compounds beforementioned and were stored in the file named “similar_comp_properties_sum.csv”.

基于计算机模拟的药物毒性及相互作用预测工作流程构建。本描述中界定之草药活性成分，系根据TSCSP数据库中特定草药所记载的自然产物，经筛选后确定，筛选标准为（OB ≥ 0.3 且 DL ≥ 0.18）。此数据集被用于对植物复合物的毒性及药物相互作用进行预测。首先，通过名为“compound_properties_mining.py”的Python 3（版本3.8.10）脚本，利用pubchempy和pandas包进行活性成分属性的挖掘，该脚本遍历“active_comp_pool_tcmsp.csv”数据集，尤其是“分子名称”列，每次提取一个“分子名称”。所收集的活性成分属性数据被写入名为“active_comp_proper_pubchem.csv”的CSV文件中。从“Toxi_infor_sum.csv”文件中筛选出的药物相互作用信息，通过名为“drug_interactions_split.py”的脚本进行分离和拆分，每行包含一个“相互作用”检索结果，以便进行进一步的人工解释。拆分后的数据存储在名为“drug_interaction_pred_0.6171.csv”的文件中。通过名为“similar_comp_crawler.py”的网页爬虫脚本，对活性成分的类似化合物进行挖掘。该脚本遍历存储活性成分属性的数据集中“活性化合物名称”列和“同分异构体SMILES”列。将同分异构体SMILES代码作为查询发送至SwissSimilarity网站（2021年12月更新的版本），选择“生物活性”化合物类别，并选择“ChEMBL（仅活性））自然产物库，基于综合方法。所有类似化合物的数据均存储在名为“similar_comp_pool_swiss.csv”的文件中。在毒性及药物相互作用信息挖掘之前，使用名为“similar_compound_properties_mining.py”的脚本，采用与上述活性成分属性挖掘相似的方法收集类似化合物的属性，并存储在名为“similar_comp_properties_sum.csv”的文件中。