摘要
利用预测系统方法,对蛋白质二级结构预测提出了一种逐步求精、多层递阶的预测系统模型,即复合金字塔模型.这种模型由4个独立协同的层面组成,通过智能接口有机融合了SAC,AAC,KDD*等源于KDTICM理论的模型和方法.模型整体贯穿物化属性与结构序列,采用因果细胞自动机选择有效物化属性,构造纯度较高的结构数据库作为训练数据源,利用领域知识与背景知识进行优化.本模型在数据集RS126及CB513分别取得83.06%与80.49%的Q3准确度,在对偏α/β型蛋白质的预测实验中,取得了93.12%的Q3准确度,并存在着进一步提高准确度的优化空间.
To attack the urgent problem in bioinformatics, protein secondary structure prediction, a gradually enhanced, multi-layered prediction systematic model, Compound Pyramid Model, is proposed. This model is composed of four independent coordination＇s layers by intelligent interfaces, synthesizes several methods, such as SVM, KDD＊ process model and so on. In this model, which intersects the amino acid phy-chemical attributes and the structural information, the effective attributes are chosen by Causal Cellular Automata, and the highly pure structure database is constructed for training. The model obtained Q3 accuracy 83.06%, 80.49% separately on data sets RS126 and CB513. And to the alpha/beta protein＇s forecast experiment, the Q3 accuracy obtained is 93.12%.
出处
《科学通报》
EI
CAS
CSCD
北大核心
2009年第21期3311-3319,共9页
Chinese Science Bulletin
基金
国家自然科学基金(批准号:69835001
60675030和60875029)
教育部科技重点项目(批准号:[2000]175)资助
关键词
蛋白质二级结构预测
复合金字塔模型
预测系统方法
数据挖掘
Compound Pyramid Model
protein secondary structure prediction
Hybrid Prediction Model
data mining
