Detection of reliable and unexpected protein fold predictions using 3D-Jury
Detection of reliable and unexpected protein fold predictions using 3D-Jury
复制标题
DOI：
10.1093/nar/gkg503
复制
发表时间：
2003-07-01
影响因子：
14.9
通讯作者：
Rychlewski, L
中科院分区：
文献类型：
--
作者：
Ginalski, K;Rychlewski, L
3D-Jury 是一种全自动蛋白质结构元预测系统，可通过 Meta Server 接口 (http://BioInfo.PL/Meta) 访问。这是元预测因子之一，对上一次 CASP-5 实验产生了巨大的、前所未有的影响。 3D-Jury 与其他元服务器相当，但它具有最高的综合特异性和敏感性。所提出的方法也非常简单且通用，甚至可以用于根据人类生成的模型集创建元预测。该系统的另一个非常重要且新颖的特征是报告的置信度得分与模型的准确性之间的高度相关性。正确预测的残基数量可以直接从预测分数估计。该方法的高可靠性使任何生物学家都可以向元服务器提交感兴趣的目标，并以相对较高的置信度进行筛选，该目标是否可以通过折叠识别方法进行预测，同时使用 PSI-Blast 等标准方法进行预测。这可以指出蛋白质或基因组注释中可能遗漏的有趣关系，并为新的科学发现提供非常有价值的信息。
3D-Jury is a fully automated protein structure meta prediction system accessible via the Meta Server interface (http://BioInfo.PL/Meta). This is one of the meta predictors, which have made a dramatic, unprecedented impact on the last CASP-5 experiment. The 3D-Jury is comparable with other meta servers but it has the highest combined specificity and sensitivity. The presented method is also very simple and versatile and can be used to create meta predictions even from sets of models produced by humans. An additional and very important and novel feature of the system is the high correlation between the reported confidence score and the accuracy of the model. The number of correctly predicted residues can be estimated directly from the prediction score. The high reliability of the method enables any biologist to submit a target of interest to the Meta Server and screen with relatively high confidence, whether the target can be predicted by fold recognition methods while being unpredictable using standard approaches like PSI-Blast. This can point to interesting relationships which could have been missed in annotations of proteins or genomes and provide very valuable information for novel scientific discoveries.