论文部分内容阅读
将一般性智能预测系统构造向结合具体领域的个性化智能预测系统构造的发展,表征了智能系统当今发展的重要趋向.蛋白质结构预测是生物信息学领域的国际性难题与主要挑战性问题之一.在本文中,我们提出了一种新型的智能预测系统模型来预测蛋白质的二级结构,它是一个多层递阶的复合金字塔模型(CPM).该模型包含4个独立的智能接口,并且综合涵盖几种相关的知识发现方法.领域知识贯穿该整个模型,并通过因果细胞自动机有效的选择属性,以及高纯度结构数据库的构造.在RS126的数据集上的预测准确性达到84.31%;在CB513数据集,预测准确性达到86.78%(居国际已知之优先地位).同时对CASP8序列预测的结果发现是优于其他方法的,如PSIPRED,Jpred,APSSP2,BehairPred等.大量实验结果表明:该智能预测系统模型具有较强的泛化能力,并提供了其他类型智能系统模型构造的方法论示范.
The development of general intelligent forecasting system architecture to the construction of personalized intelligent forecasting system which combines with specific areas represents an important tendency of the development of intelligent system today.Protein structure prediction is one of the international and the main challenge in the field of bioinformatics In this paper, we propose a novel intelligent predictive system model to predict the protein secondary structure, which is a multi-level hierarchical compound pyramid model (CPM) that contains four independent intelligent interfaces and It comprehensively covers several related knowledge discovery methods.The domain knowledge runs through the whole model, and through the causal cell automata effective selection attributes and the construction of high purity structure database, the prediction accuracy on the RS126 dataset reaches 84.31%. In the CB513 dataset, the prediction accuracy reached 86.78%, which is internationally known as the priority.At the same time, the CASP8 sequence prediction results were found to be superior to other methods, such as PSIPRED, Jpred, APSSP2, BehairPred, etc. A large number of experimental results show that: The intelligent forecasting system model has strong generalization ability and provides the methodological demonstration of other types of intelligent system model construction .