Latek Dorota, Bajda Marek, Filipek Sławomir
Faculty of Chemistry, Biological and Chemical Research Centre, University of Warsaw , Pasteura 1, 02-093 Warsaw, Poland.
Department of Physicochemical Drug Analysis, Faculty of Pharmacy, Medical College, Jagiellonian University , Medyczna 9, 30-688 Cracow, Poland.
J Chem Inf Model. 2016 Apr 25;56(4):630-41. doi: 10.1021/acs.jcim.5b00451. Epub 2016 Mar 29.
The recent GPCR Dock 2013 assessment of serotonin receptor 5-HT1B and 5-HT2B, and smoothened receptor SMO targets, exposed the strengths and weaknesses of the currently used computational approaches. The test cases of 5-HT1B and 5-HT2B demonstrated that both the receptor structure and the ligand binding mode can be predicted with the atomic-detail accuracy, as long as the target-template sequence similarity is relatively high. On the other hand, the observation of a low target-template sequence similarity, e.g., between SMO from the frizzled GPCR family and members of the rhodopsin family, hampers the GPCR structure prediction and ligand docking. Indeed, in GPCR Dock 2013, accurate prediction of the SMO target was still beyond the capabilities of most research groups. Another bottleneck in the current GPCR research, as demonstrated by the 5-HT2B target, is the reliable prediction of global conformational changes induced by activation of GPCRs. In this work, we report details of our protocol used during GPCR Dock 2013. Our structure prediction and ligand docking protocol was especially successful in the case of 5-HT1B and 5-HT2B-ergotamine complexes for which we provide one of the most accurate predictions. In addition to a description of the GPCR Dock 2013 results, we propose a novel hybrid computational methodology to improve GPCR structure and function prediction. This computational methodology employs two separate rankings for filtering GPCR models. The first ranking is ligand-based while the second is based on the scoring scheme of the recently published BCL method. In this work, we prove that the use of knowledge-based potentials implemented in BCL is an efficient way to cope with major bottlenecks in the GPCR structure prediction. Thereby, we also demonstrate that the knowledge-based potentials for membrane proteins were significantly improved, because of the recent surge in available experimental structures.
近期GPCR Dock 2013对5-羟色胺受体5-HT1B和5-HT2B以及smoothened受体SMO靶点的评估，揭示了当前所使用计算方法的优缺点。5-HT1B和5-HT2B的测试案例表明，只要目标-模板序列相似度相对较高，受体结构和配体结合模式就能以原子细节精度进行预测。另一方面，低目标-模板序列相似度的情况，例如卷曲GPCR家族的SMO与视紫红质家族成员之间，会妨碍GPCR结构预测和配体对接。事实上，在GPCR Dock 2013中，对SMO靶点的准确预测仍超出大多数研究团队的能力范围。如5-HT2B靶点所示，当前GPCR研究中的另一个瓶颈是对GPCR激活诱导的全局构象变化的可靠预测。在这项工作中，我们报告了在GPCR Dock 2013期间使用的方案细节。我们的结构预测和配体对接方案在5-HT1B和5-HT2B-麦角胺复合物的案例中特别成功，我们为此提供了最准确的预测之一。除了描述GPCR Dock 2013的结果外，我们还提出了一种新颖的混合计算方法来改进GPCR结构和功能预测。这种计算方法采用两种单独的排名来筛选GPCR模型。第一种排名基于配体，而第二种基于最近发表的BCL方法的评分方案。在这项工作中，我们证明了使用BCL中实现的基于知识的势是应对GPCR结构预测中主要瓶颈的有效方法。由此，我们还证明了由于最近可用实验结构的激增，膜蛋白的基于知识的势得到了显著改进。
