基于组合的序列特征识别蛋白质复杂超二级结构βαβ模体

Recognition of Complex Super Secondary Structure βαβ motifs in Proteins based on Combined sequence features

从已知蛋白质结构的氨基酸序列出发,利用DSSP和PROMTIF构建了蛋白质复杂超二级结构strand-loop-helix-loop-strand模体数据集。数据集含1458条蛋白质链,其中βαβ模体数为3632个,非βαβ模体数为3148个。将亲疏水组分、优化的位点氨基酸组分、预测的模体信息和二级结构信息共同作为序列特征输入支持向量机,5交叉检验的预测总精度和马氏相关系数达到了79.7%和0.59;独立检验的预测总精度和马氏相关系数达到了73.4%和0.47。

From the amino acid sequence of the known protein structure,we constructed complex secondary structureβαβmotifs datasets by using the DSSP and PROMTIF.Then the core structure,loop-helix-loop,of βαβ motif was analyzed,and the research object that loop-helix-loop length is from 10 to 26amino acids was selected.The dataset contained 1458 proteins,among 3632βαβmotifs and 3148non-βαβmotifs.Here we applied hydropathy,optimized amino acid composition of position,predicted function motif and predicted secondary structure information as combined sequence feature and input SVM algorithm.The overall accuracy and Matthew＇s correlation coefficient of 5-fold cross-validation achieved 79.7% and 0.59.The overall accuracy and Matthew＇s correlation coefficient of independent test achieved 73.4% and 0.47.