Cohen Serge X, Ben Jelloul Marouane, Long Fei, Vagin Alexei, Knipscheer Puck, Lebbink Joyce, Sixma Titia K, Lamzin Victor S, Murshudov Garib N, Perrakis Anastassis
Department of Molecular Carcinogenesis, Netherlands Cancer Institute, Amsterdam, The Netherlands.
Acta Crystallogr D Biol Crystallogr. 2008 Jan;64(Pt 1):49-60. doi: 10.1107/S0907444907047580. Epub 2007 Dec 5.
Automatic iterative model (re-)building, as implemented in ARP/wARP and its new control system flex-wARP, is particularly well suited to follow structure solution by molecular replacement. More than 100 molecular-replacement solutions automatically solved by the BALBES software were submitted to three standard protocols in flex-wARP and the results were compared with final models from the PDB. Standard metrics were gathered in a systematic way and enabled the drawing of statistical conclusions on the advantages of each protocol. Based on this analysis, an empirical estimator was proposed that predicts how good the final model produced by flex-wARP is likely to be based on the experimental data and the quality of the molecular-replacement solution. To introduce the differences between the three flex-wARP protocols (keeping the complete search model, converting it to atomic coordinates but ignoring atom identities or using the electron-density map calculated from the molecular-replacement solution), two examples are also discussed in detail, focusing on the evolution of the models during iterative rebuilding. This highlights the diversity of paths that the flex-wARP control system can employ to reach a nearly complete and accurate model while actually starting from the same initial information.
自动迭代模型（重新）构建，如ARP/wARP及其新的控制系统flex-wARP中所实现的，特别适合于通过分子置换来追踪结构解析过程。由BALBES软件自动解析的100多个分子置换解决方案被提交到flex-wARP中的三种标准协议中，并将结果与来自蛋白质数据库（PDB）的最终模型进行比较。标准指标以系统的方式收集，并能够就每种协议的优势得出统计结论。基于该分析，提出了一种经验估计器，该估计器可根据实验数据和分子置换解决方案的质量预测flex-wARP生成的最终模型可能有多好。为了介绍flex-wARP的三种协议之间的差异（保留完整的搜索模型、将其转换为原子坐标但忽略原子身份或使用根据分子置换解决方案计算的电子密度图），还详细讨论了两个例子，重点关注迭代重建过程中模型的演变。这突出了flex-wARP控制系统在实际上从相同的初始信息开始时，可以采用多种路径来获得一个近乎完整且准确的模型。
