摘要
氨基酸描述子是采用多元统计方法从大量氨基酸性质参数中提取得到的少数关键信息成分,目前已被广泛应用于多肽生物活性预测及蛋白质功能判别等领域。鉴于近年来氨基酸描述子种类的极速扩增,收集了目前国内外发表的27种氨基酸描述子,并将其用于8组经典多肽集的结构表征及定量构效关系研究。通过系统比较这些描述子对同一肽集与不同肽集的统计建模结果,我们认为物理化学描述子的建模效果优于拓扑描述子,拓扑描述子的建模效果优于三维结构描述子,且已有诸多氨基酸描述子已经达到性能限度,如未考虑肽链内部各氨基酸残基的交互影响,肽配基与相应靶标蛋白的相互结合,因此不再建议按照传统思路进一步提出新型描述子种类。
Amino acid descriptors are key information extracted from a large number of amino acid properties by multivariate statistical methods. Recently, they have been widely applied to predicting the biological activities of polypeptides and identifying the function of a protein. In view of the rapid expansion of amino acid descriptors in recent years, in this paper, we collected 27 amino acid descrip- tors published at home and abroad. These descriptors were then employed to explore the structural and functional relationship of 8 groups of classical polypeptides. By comparing results of modeling statistic in the same and different sets of peptides, we conclude that physicochemical descriptor model is better than topological descriptor model, and topological descriptor model is better than 3D struc- tural descriptor model. Many descriptors of amino acid have reached the performance limits, therefore, we do not suggest proposing new descriptor of amino acid according to traditional ideas.
出处
《生物学杂志》
CAS
CSCD
2014年第1期87-93,共7页
Journal of Biology
基金
中央高校基本科研业务费科技创新项目资助(编号:SWJTU11CX113)
