Pharmaceutical Data Exploration Laboratory, Department of Pharmacy, National University of Singapore, Singapore 117543, Singapore.
Mol Divers. 2012 Aug;16(3):513-24. doi: 10.1007/s11030-012-9384-z. Epub 2012 Jul 21.
Due to the importance of neuraminidase in the pathogenesis of influenza virus infection, it has been regarded as the most important drug target for the treatment of influenza. Resistance to currently available drugs and new findings related to structure of the protein requires novel neuraminidase 1 (N1) inhibitors. In this study, a consensus QSAR model with defined applicability domain (AD) was developed using published N1 inhibitors. The consensus model was validated using an external validation set. The model achieved high sensitivity, specificity, and overall accuracy along with low false positive rate (FPR) and false discovery rate (FDR). The performance of model on the external validation set and training set were comparable, thus it was unlikely to be overfitted. The low FPR and low FDR will increase its accuracy in screening large chemical libraries. Screening of ZINC library resulted in 64,772 compounds as probable N1 inhibitors, while 173,674 compounds were defined to be outside the AD of the consensus model. The advantage of the current model is that it was developed using a large and diverse dataset and has a defined AD which prevents its use on compounds that it is not capable of predicting. The consensus model developed in this study is made available via the free software, PaDEL-DDPredictor.
由于神经氨酸酶在流感病毒感染发病机制中的重要性，它一直被视为治疗流感的最重要的药物靶点。对现有药物的耐药性和与蛋白质结构相关的新发现要求开发新的神经氨酸酶 1（N1）抑制剂。在这项研究中，使用已发表的 N1 抑制剂开发了具有明确适用域（AD）的共识 QSAR 模型。使用外部验证集对共识模型进行了验证。该模型具有较高的灵敏度、特异性和总体准确性，以及较低的假阳性率（FPR）和假阴性率（FDR）。该模型在外部验证集和训练集上的性能相当，因此不太可能过拟合。低 FPR 和低 FDR 将提高其在筛选大型化学库中的准确性。对 ZINC 库的筛选得到了 64772 种可能的 N1 抑制剂，而有 173674 种化合物被定义为共识模型适用域之外的化合物。当前模型的优势在于，它是使用大型和多样化的数据集开发的，并且具有明确的适用域，这可以防止在它无法预测的化合物上使用。本研究中开发的共识模型可通过免费软件 PaDEL-DDPredictor 使用。
