First report on retention time prediction of pesticides and veterinary drugs in cow milk using read-across and intelligent consensus prediction: an alternative for hazard assessment employing food-informatics

Milk is one of the primary sources of food. Pesticides and veterinary drugs are reaching directly or indirectly (pesticides containing grass or other cattle foods) into the milk of the cattle, which are serious health concerns to the animals, infants, babies, and humans. So, in-silico approaches like QSPR, read-across, etc., are used as an alternative (reduce time, cost, complex analytical process) for calculating retention time (RT). The present work involves the development of the first multiple PLS-based QSAR models for the estimation of RT of pesticides, veterinary drugs, and related chemical hazards in milk by strictly obeying the OECD principles. Based on the results, the quality of the models is good enough. In the current work, it was observed that lipophilicity, binding property, rotatable bonds, and reactivity are responsible for high RT while hydrophilicity, the presence of primary amines, aqueous solubility, and branching reduce the RT of the compounds. The established models were utilized to screen the PPDB database to justify its real-world application. The present study will be vital in the food-informatics area for the RT data-gap filling and identification of hazardous chemicals in milk. Thus, it will be helpful to maintain a healthier, safer, and eco-friendly ecosystem.

牛乳是人类主要的食物来源之一。农药与兽药可通过直接或间接途径（如喷施于牧草或其他牛用饲料的农药）进入牛乳中，对畜禽、婴幼儿及人类的健康构成严重威胁。因此，诸如定量结构-保留相关性（Quantitative Structure-Retention Relationship, QSPR）、交叉参照法等计算机虚拟方法，被用作计算保留时间（Retention Time, RT）的替代方案，可有效缩减实验耗时、降低成本并简化复杂的分析流程。本研究严格遵循经济合作与发展组织（Organisation for Economic Co-operation and Development, OECD）原则，构建了首个基于偏最小二乘（Partial Least Squares, PLS）的多变量定量结构-活性相关性（Quantitative Structure-Activity Relationship, QSAR）模型，用于预测牛乳中农药、兽药及相关化学危害物的保留时间。经检验，所建模型具备优异的预测性能。本研究发现，亲脂性、结合特性、可旋转键数目与反应活性会延长化合物的保留时间，而亲水性、伯胺基团的存在、水溶性及分子支化度则会缩短化合物的保留时间。所建模型被用于筛选PPDB数据库，以验证其实际应用价值。本研究可为食品信息学领域填补牛乳中保留时间数据的空白，并助力牛乳中有害化学物质的识别，具备重要的应用价值，进而有助于构建更健康、更安全且环境友好的生态系统。