融合物化特征及结构信息的随机森林算法识别Ca^(2+),Mg^(2+)和Mn^(2+)结合残基

Identification Ca^(2+),Mg^(2+) and Mn^(2+) Binding Residue Based on Random Forest Algorithm Using Physical and Chemical Characteristics and Structure Information

Ca^(2+),Mg^(2+)和Mn^(2+)结合残基的识别是众多的金属离子配体中效果较差的,但这些金属离子配体与蛋白质的相互作用在生命进程中实现了不可替代的生物学功能,因此提高Ca^(2+),Mg^(2+)和Mn^(2+)结合残基的识别质量对药物设计特别重要。工作中提出了一种融合物化特征及结构信息的随机森林预测方法,对Ca^(2+),Mg^(2+)和Mn^(2+)配体结合残基预测的总精度分别达到了77.6%,77.0%和87.1%,马氏相关系数分别达到了0.563,0.542和0.743。同时将同样的预测参数降维后应用于SVM方法对三种金属离子结合残基进行了预测,也得到了较好的预测结果。

The identification effect of Ca2+,Mg2+and Mn2+binding residues are relatively poor in metal ions ligand researches,but the interaction between protein and these metal ions can realize irreplaceable biological function in life processes.Therefore improving the Ca2+,Mg2+and Mn2+binding residue identification is particularly important for drugs design.In this paper,a random forest algorithm with fusion of physical and chemical characteristics and structure information is proposed.For Ca2+,Mg2+and Mn2+ligand binding residues,the prediction accuracy respectively achieved 77.6%,77.0%and 87.1%,and the prediction MCC respectively reached 0.563,0.542 and 0.743.In addition,we used SVM algorithm with same predictive parameters to predict binding residues.As a result,better results are obtained.