基于知识的蛋白质结构预测评分函数的研究-生物医学工程专业论文.docxVIP

布的评分函数。并通过多次实验确定计算距离分布时的离散区间数目为20。 布的评分函数。并通过多次实验确定计算距离分布时的离散区间数目为 20。 (2)在蛋白质结构中，主链二面角(≯，妒)的分布就可用拉氏构象 图来描画。本文构建了一个基于二面角的评分函数，通过计算确定把 (≯，伊)空间离散为60的网格是最好的选择。 (3)进一步组合上述从距离和角度两个方面建立的评分函数，所得 的评分函数性能比前两者有大幅提高。通过正确识别蛋白质天然结构总 数和Z score这两个性能指标，确定了性能最好的一组组合能量，此组 合能量函数能识别出150条天然结构的测试集中的109条。 (4)由于20种氨基酸在蛋白质中出现的频率不一样，因此存在着 数据稀疏性。本文采用了一种稀疏数据校正策略，通过计算确定了另一 组识别性能最优的组合能量，能识别1 14条天然结构，识别率为76％， Z_score值也同时得到改善。 关键词：基于知识的评分函数，能量函数，依赖距离分布的能量函数， 二面角能量函数，组合能量函数，稀疏数据校正 A A KNOWLEDGE—BASED SCOIUNG FUNCTION FOR PREDICTING PROTEIN STRUCTURES ABSTRACT It has recently been a challenging research topic in bioinformatics to predict the tertiary structure of a protein from its amino acid sequence．It is critical to design a good scoring function in recognizing the native structure of a protein． Scoring function，which is also called as energy function or potential function，can be classified into two categories：Physics—based scoring function and knowledge-based scoring function．The former is all experiential formula resulting from analyzing the forces between the particles SO it really reflects the forces between the particles inside the protein or between the particles of the protein and solvent．It is complicated and very time—consuming to calculate physics—based scoring function．The latter is derived from the known protein structures data in protein database(PDB)as training data and is statistically effective．The knowledge-based scoring function implicitly represents the physical and chemical forces in the native protein structures and can be computed more easily than the former．The performance of knowledge—based scoring function largely depends on the quantity and quality of known protein structures that were selected as the training data． In this paper,we employed proteins listed in pdb select 25 as learning data．As the selection ofa representative set ofPDB chains，the pdb select 25 list has been continually updated by the member of European Molecular Biology Biology Laboratory．The main p