Evaluation of designed ligands by a multiple screening method: Application to glycogen phosphorylase inhibitors constructed with a variety of approaches

Glycogen phosphorylase (GP) is an important enzyme that regulates blood glucose level and a key therapeutic target for the treatment of type II diabetes. In this study, a number of potential GP inhibitors are designed with a variety of computational approaches. They include the applications of MCSS, LUDI and CoMFA to identify additional fragments that can be attached to existing lead molecules; the use of 2D and 3D similarity-based QSAR models (HQSAR and SMGNN) and of the LUDI program to identify novel molecules that may bind to the glucose binding site. The designed ligands are evaluated by a multiple screening method, which is a combination of commercial and in-house ligand-receptor binding affinity prediction programs used in a previous study (So and Karplus, J. Comp.-Aid. Mol. Des., 13 (1999), 243–258). Each method is used at an appropriate point in the screening, as determined by both the accuracy of the calculations and the computational cost. A comparison of the strengths and weaknesses of the ligand design approaches is made.     

Evaluation of binary QSAR models derived from LUDI and MOE scoring functions for structure based virtual screening

In today's world of high-throughput in silico screening, the development of virtual screening methodologies to prioritize small molecules as new chemical entities (NCEs) for synthesis is of current interest. Among several approaches to virtual screening, structure-based virtual screening has been considered the most effective. However the problems associated with the ranking of potential solutions in terms of scoring functions remains one of the major bottlenecks in structure-based virtual screening technology. It has been suggested that scoring functions may be used as filters for distinguishing binders from nonbinders instead of accurately predicting their binding free energies. Subsequently, several improvements have been made in this area, which include the use of multiple rather than single scoring functions and application of either consensus or multivariate statistical methods or both to improve the discrimination between binders and nonbinders. In view of it, the discriminative ability (distinguishing binders from nonbinders) of binary QSAR models derived using LUDI and MOE scoring functions has been compared with the models derived by Jacobbsson et al. on five data sets viz. estrogen receptor alphamimics (ERalpha_mimics), estrogen receptor alphatoxins (ERalpha_toxins), matrix metalloprotease 3 inhibitors (MMP-3), factor Xa inhibitors (fXa), and acetylcholine esterase inhibitors (AChE). The overall analyses reveal that binary QSAR is comparable to the PLS discriminant analysis, rule-based, and Bayesian classification methods used by Jacobsson et al. Further the scoring functions implemented in LUDI and MOE can score a wide range of protein-ligand interactions and are comparable to the scoring functions implemented in ICM and Cscore. Thus the binary QSAR models derived using LUDI and MOE scoring functions may be useful as a preliminary screening layer in a multilayered virtual screening paradigm.     

基于机器学习方法的丙型肝炎病毒聚合酶NS5B非核苷抑制剂的定量构效关系研究

对89 个苯并异噻唑和苯并噻嗪类丙型肝炎病毒(HCV) NS5B聚合酶非核苷抑制剂进行了定量构效关系(QSAR)研究. 采用遗传算法组合偏最小二乘(GA-PLS)和线性逐步回归分析(LSRA)两种特征选择方法选择最优描述符子集, 然后建立多元线性回归和偏最小二乘线性回归模型. 并首次尝试使用遗传算法耦合支持向量机方法(GA-SVM)对两种特征选择方法所选的描述符子集分别建立非线性支持向量机回归模型. 三种机器学习方法所建模型均得到比较满意的预测效果. 采用LSRA所选的6 个描述符建立的三个QSAR模型对于测试集的相关系数为0.958-0.962, GA-SVM法给出最好的预测精度(0.962). 采用GA-PLS所选的7个描述符建立的三个QSAR模型对于测试集的相关系数为0.918-0.960, 偏最小二乘回归模型的结果最好(0.960). 本工作提供了一种有效的方法来预测丙型肝炎病毒抑制剂的生物活性, 该方法也可以扩展到其他类似的定量构效关系研究领域.     

Three-dimensional quantitative structure-activity and structure-selectivity relationships of dihydrofolate reductase inhibitors

Three-dimensional quantitative structure-activity relationship (3D-QSAR) modelling using comparative molecular similarity indices analysis (CoMSIA) was applied to a series of 406 structurally diverse dihydrofolate reductase (DHFR) inhibitors from Pneumocystis carinii (pc) and rat liver (rl). X-ray crystal structures of three inhibitors bound to pcDHFR were used for defining the alignment rule. For pcDHFR, a QSAR model containing 6 components was selected using leave-10%-out cross-validation (n= 240, q2 = 0.65), while a 4-component model was selected for rlDHFR (n= 237, q2 = 0.63); both include steric, electrostatic and hydrophobic contributions. The models were validated using a large test set, designed to maximise its diversity and to verify the predictive accuracy of models for extrapolation. The pcDHFR model has r2 = 0.60 and mean absolute error (MAE) = 0.57 for the test set after removing 4 outliers, and the rlDHFR model has r2 = 0.60 and MAE = 0.69 after removing 4 test set outliers. In addition, classification models predicting selectivity for pcDHFR over rlDHFR were developed using soft independent modelling by class analogy (SIMCA), with a selectivity ratio of 2 (IC50,rlDHFR/ IC50,pcDHFR) used for delimiting classes. A 5-component model including steric and electrostatic contributions has cross-validated and test set classification rates of 0.67 and 0.68 for selective inhibitors, and 0.85 and 0.72 for unselective inhibitors. The predictive accuracy of models, together with the identification of important contributions in QSAR and classification models, offer the possibility of designing potent selective inhibitors and estimating their activity prior to synthesis.