The authors have proposed an efficient multilevel prediction model for better activity assessment to test whether certain chemical compounds can disrupt processes in the human body that may create negative health effects. Here, a computational method (in-silico) is proposed for the quality prediction of drugs in terms of their activity, activity score, potency, and efficacy for estrogen receptors (ERs) by using various physicochemical properties (molecular descriptors). PaDEL-Descriptor is used for features extraction. The ER dataset has 8481 drug molecules where 1084 are active, and 7397 are inactive, and each drug molecule has 1444 features. This dataset is highly imbalanced and has a substantial number of features. Initially, a class imbalance problem is resolved through synthetic minority oversampling technique algorithm, and feature selection is done using FSelector library of R. A machine learning based multilevel prediction model is developed where classification is performed on its first level and regression on its second level. By using all these strategies simultaneously, outperformed accuracy is achieved in comparison to many other computational approaches. The K-fold cross-validation is performed to measure the consistency of the model for all the target classes. Finally, the validity of the proposed method on some AIDS therapy's drug molecules is proved.

中文翻译：

---

使用多级预测模型评估雌激素受体中小药物分子的活性。

作者提出了一种有效的多级预测模型，以更好地评估活动，以测试某些化合物是否会扰乱人体过程，从而对健康产生负面影响。在这里，提出了一种计算方法（计算机），通过使用各种物理化学特性（分子描述符）来预测药物的活性、活性评分、效力和对雌激素受体（ER）的功效。PaDEL-Descriptor用于特征提取。ER 数据集有 8481 个药物分子，其中 1084 个是活性的，7397 个是非活性的，每个药物分子有 1444 个特征。该数据集高度不平衡，并且具有大量特征。首先，通过合成少数过采样技术算法解决类不平衡问题，并使用 R 的 FSelector 库完成特征选择。开发了基于机器学习的多级预测模型，其中在第一级执行分类，在第二级执行回归。通过同时使用所有这些策略，与许多其他计算方法相比，可以实现更高的准确性。执行 K 折交叉验证是为了衡量所有目标类别的模型的一致性。最后，证明了该方法对一些艾滋病治疗药物分子的有效性。