摘要
基于模板的模建方法是蛋白质结构预测领域中最为准确有效的方法,该类方法的成功与否对模板质量的要求较高。为待预测序列找寻合适的模板,本文提出了一种profile-profile比对的方法将查询序列同模板库中的已知结构蛋白进行比对,然后根据比对结果的Z-score得分高低顺序挑选出合适的模板。结果表明:本文的profile-profile比对方法在测试集上的性能明显优于PSI-BLAST,相比PSI-BLAST在测试集上的准确度提高了约14.3%,配对t检验的结果表明准确度的提高具有统计显著性。从而得出如下结论:本文的profile-profile比对方法可以用于为序列相似性较低的待预测序列搜索远距离同源模板,并用于指导后续的三级结构预测。
Template - based modeling was the most accurate and efficient method in the field of protein tertiary structure prediction, however, this Kind of method was highly dependent on the template quality. This article was aimed at establishing a kind of alignment method, which was used for detection of suitable templates for query se- quence. A kind of profile - profile alignment method was proposed in this article, which firstly make alignments of query sequences and proteins with known structure in template library, followed the templates being selected out ac- cording to Z - score ranking of alignment results. The proposed profile - profile alignment method obviously outper- formed PSI - BLAST on the testing sets. The accuracy was increased by 14.3% in comparison to PSI - BLAST on testing sets, which was statistically significant on a paired Students＇ t test. The profile - profile alignment method in this article could be employed to identify distantly -related templates for query sequences with low sequence simi- larity, furthermore, the templates obtained from this method could be used to guide tertiary structure prediction of query sequence subsequently.
出处
《生物信息学》
2013年第1期16-21,共6页
Chinese Journal of Bioinformatics
