摘要
针对蛋白质关联图预测问题,提出一种克隆选择算法与蛋白质折叠规律相结合的预测方法,综合使用蛋白质序列疏水性质、残基的二级结构倾向、关联图总点数等信息,构造了基于限制规则的克隆选择算法适应度函数,设计了符合关联图生物学特性的变异操作。算法不需要使用额外蛋白质作为训练集,不需要从现有蛋白质数据库中提取模板,因此不受现有蛋白质结构数据的局限,可以由序列信息直接进行预测。对200个非同源蛋白质的测试验证了算法的有效性。
An improved clonal selection algorithm for protein contact map prediction is proposed. The fitness function of the algorithm is constructed by using the protein folding restrictions, such as the hydrophobicity of amino acids, the secondary structure inclination of residues, the total number of contacts in contact map, and so on. Variance operation according with the biological properties of the contact map is designed. This algorithm does not need additional proteins as the training set to extract the template; therefore it is not affected by the existing limitations of protein structure. Prediction tests for 200 non-homological proteins with different lengths are conducted and the results verify the effectiveness of the algorithm.
出处
《吉林大学学报(工学版)》
EI
CAS
CSCD
北大核心
2009年第5期1303-1308,共6页
Journal of Jilin University:Engineering and Technology Edition
基金
'863'国家高技术研究发展计划项目(2007AA04Z114)
国家自然科学基金项目(60673099
60873146)
关键词
人工智能
蛋白质关联图
免疫算法
克隆选择算法
疏水性
artificial intelligence
protein contact maps
immune algorithm
clonal selection algorithm
amino acids hydrophobicity