ABSTRACT

Robust and reliable QSAR models were developed to predict half-maximal inhibitory concentration (IC₅₀) values of hepatitis C virus NS3/4A protease inhibitors from the Monte Carlo technique. 524 HCV NS3/4A protease inhibitors were extracted from the scientific literature to create a reasonably large set. The models were developed using CORAL software by using two target functions namely target function 1 (TF1) without applying the index of ideality of correlation (IIC) and target function 2 (TF2) that uses IIC. The constructed models based on TF2 were statistically more significant and robust than the models based on TF1. The determination coefficients (r²) of training and test sets were 0.86 and 0.88 for the best split based on TF2. The promoters of the increase/decrease of activity were also extracted and interpreted in detail. The model interpretation results explain the role of different structural attributes in predicting the pIC₅₀ values of hepatitis C virus NS3/4A protease inhibitors. Based on the mechanistic model interpretation results, eight new compounds were designed and their pIC₅₀ values were predicted based on the average prediction of ten models.

摘要

开发了稳健可靠的 QSAR 模型来预测来自蒙特卡罗技术的丙型肝炎病毒 NS3/4A 蛋白酶抑制剂的半数最大抑制浓度 (IC ₅₀ ) 值。从科学文献中提取了 524 种 HCV NS3/4A 蛋白酶抑制剂，以创建一个相当大的集合。这些模型是使用 CORAL 软件通过使用两个目标函数开发的，即目标函数 1 (TF1) 不应用相关性理想指数 (IIC) 和使用 IIC 的目标函数 2 (TF2)。基于 TF2 构建的模型比基于 TF1 的模型在统计上更显着和稳健。决定系数 ( r ²) 的训练集和测试集分别为 0.86 和 0.88，用于基于 TF2 的最佳分割。还提取并详细解释了活性增加/减少的启动子。模型解释结果解释了不同结构属性在预测丙型肝炎病毒 NS3/4A 蛋白酶抑制剂pIC ₅₀值中的作用。根据机理模型解释结果，设计了八种新化合物，并根据十个模型的平均预测值预测了它们的 pIC ₅₀值。