Critical assessment of methods of protein structure prediction (CASP): Round IV
Critical assessment of methods of protein structure prediction (CASP): Round IV
复制标题
蛋白质结构预测 (CASP) 方法的批判性评估：第四轮
DOI：
10.1002/prot.10054
复制
发表时间：
2001
期刊：
影响因子：
--
通讯作者：
T. Hubbard
中科院分区：
文献类型：
--
作者：
J. Moult;K. Fidelis;A. Zemla;T. Hubbard
This issue of Proteins is devoted to articles reporting the outcome of the fourth communitywide experiment to assess methods of protein structure prediction in CASP4. Methods are assessed on the basis of the analysis of a large number of blind predictions of protein structure. Three previous experiments, in 1994, 1996 and 1998, were reported in previous special issues of Proteins and elsewhere. Several independent discussions of CASP4 have also appeared. Early experiments focused on establishing what then-current methods could or could not deliver. Later experiments have extended the significance of the earlier results by including a larger number of predictions from more investigators, and by measuring the extent to which there has been progress in each of the prediction areas during the intervening 2 years. With a series of four experiments conducted over a period of 8 years, it has become clear where the bottlenecks to progress are, where progress is being made, and at what sort of rate. This article outlines the structure and conduct of the experiment. Then there is a description of the CASP4 target proteins. The following article describes some of the mechanics of the process and the newer methods for numerical evaluation of predictions. The bulk of this issue consists of three sections, one for each area of prediction: comparative modeling, fold recognition, and “new fold” methods. Each section contains an article by the assessment team in that area, as well as contributions from the prediction groups the assessors considered to have done the most interesting work. The number of predictor articles in each category vary, with only two in comparative modeling, but four in the fold recognition and six in the new folds category. The small number of articles in comparative modeling reflects the fact that there appears to have been little progress in this area since the last CASP. The assessors’ articles are probably the most important in the issue, describing the state of the art as they found it in CASP4. As in CASP3, results are also reported for a parallel experiment, CAFASP (Critical assessment of fully automatic prediction methods). CAFASP2 makes use of the CASP target distribution and prediction collection infrastructure but is otherwise independent. The goal of CAFASP is to assess the state of the art in automatic methods of structure prediction. Whereas CASP allows any combination of computational and human methods, CAFASP captures predictions directly from fully automatic servers. CAFASP results were also evaluated by the CASP assessors, providing a comparison of fully automatic and hybrid methods. The most significant comparison is provided in the fold recognition category. A change in CASP4 is the inclusion of large-scale benchmarking of prediction methods. Two of these methods are described: EVA and LiveBench. These experiments complement CASP, particularly by clarifying issues of the statistical significance of the results. Both operate by sending just-released Protein Data Bank (PDB) entries to automatic prediction servers and by collating and analyzing the results over time. LiveBench focuses primarily on the area of fold recognition, and EVA primarily on secondary structure predictions. We expect that well-organized benchmarking experiments will play an increasing role in the future of prediction assessment. Finally, this issue contains an article that compares performance in the different CASPs in some areas.
