Optimal selection of suitable templates in protein interface prediction.
Optimal selection of suitable templates in protein interface prediction.
复制标题
DOI：
10.1093/bioinformatics/btad510
复制
发表时间：
2023-09-02
期刊：
影响因子：
--
通讯作者：
中科院分区：
文献类型：
--
作者：
蛋白质-蛋白质界面的分子水平分类可以极大地帮助功能表征和合理的药物设计。最准确的蛋白质界面预测依赖于寻找具有已知界面的同源蛋白质，因为大多数界面在同一蛋白质家族中是保守的。这些基于模板的预测方法的准确性取决于合适模板的正确选择。在免疫球蛋白超家族 (IgSF) 中选择正确的模板具有挑战性，因为其成员序列同一性较低，并且尽管结构同源，但仍显示出广泛的替代结合位点。我们提出了一种预测蛋白质界面的新方法。首先，使用基于相互信息的方法建立特定于模板的、信息丰富的进化概况。接下来，基于源自进化谱的残基水平保守分数的相似性，查询蛋白质与其超家族中具有已知接口定义的所有可用模板蛋白质进行分层聚类。一旦聚类，就会选择最密切相关的模板的子集，并进行界面预测。这些最初的界面预测随后通过广泛的对接进行完善。该方法以 51 种 IgSF 蛋白为基准，可以预测 IgSF 蛋白的非平凡界面，平均 F 分数和中值 F 分数分别为 0.64 和 0.78。我们还提供了一种评估结果置信度的方法。如果删除 27% 的低置信度案例和 17% 的中等置信度案例，则平均 F 分数和中位 F 分数分别增加到 0.8 和 0.81。最后，我们提供残基水平界面预测、蛋白质复合物和 IgSF 中单个个体的置信度测量。源代码可免费获取：https://gitlab.com/fiserlab.org/interdct_with_refinement。
Molecular-level classification of protein–protein interfaces can greatly assist in functional characterization and rational drug design. The most accurate protein interface predictions rely on finding homologous proteins with known interfaces since most interfaces are conserved within the same protein family. The accuracy of these template-based prediction approaches depends on the correct choice of suitable templates. Choosing the right templates in the immunoglobulin superfamily (IgSF) is challenging because its members share low sequence identity and display a wide range of alternative binding sites despite structural homology. We present a new approach to predict protein interfaces. First, template-specific, informative evolutionary profiles are established using a mutual information-based approach. Next, based on the similarity of residue level conservation scores derived from the evolutionary profiles, a query protein is hierarchically clustered with all available template proteins in its superfamily with known interface definitions. Once clustered, a subset of the most closely related templates is selected, and an interface prediction is made. These initial interface predictions are subsequently refined by extensive docking. This method was benchmarked on 51 IgSF proteins and can predict nontrivial interfaces of IgSF proteins with an average and median F-score of 0.64 and 0.78, respectively. We also provide a way to assess the confidence of the results. The average and median F-scores increase to 0.8 and 0.81, respectively, if 27% of low confidence cases and 17% of medium confidence cases are removed. Lastly, we provide residue level interface predictions, protein complexes, and confidence measurements for singletons in the IgSF. Source code is freely available at: https://gitlab.com/fiserlab.org/interdct_with_refinement.
登录
查看更多内容
DOI：
10.1007/978-1-60761-842-3_6
发表时间：
2010
期刊：
Methods in molecular biology (Clifton, N.J.)
影响因子：
--
作者：
Fiser A
通讯作者：
Fiser A
影响因子：
64.8
作者：
Jumper J;Evans R;Pritzel A;Green T;Figurnov M;Ronneberger O;Tunyasuvunakool K;Bates R;Žídek A;Potapenko A;Bridgland A;Meyer C;Kohl SAA;Ballard AJ;Cowie A;Romera-Paredes B;Nikolov S;Jain R;Adler J;Back T;Petersen S;Reiman D;Clancy E;Zielinski M;Steinegger M;Pacholska M;Berghammer T;Bodenstein S;Silver D;Vinyals O;Senior AW;Kavukcuoglu K;Kohli P;Hassabis D
通讯作者：
Hassabis D
影响因子：
5.8
作者：
Murakami, Yoichi;Mizuguchi, Kenji
通讯作者：
Mizuguchi, Kenji
影响因子：
5.8
作者：
Gil, Nelson;Fiser, Andras
通讯作者：
Fiser, Andras
影响因子：
5.8
作者：
Gil，Nelson;Fiser，Andras
通讯作者：
Fiser，Andras