基于常用得分矩阵的神经网络法预测蛋白质的二级结构_徐建平.pdf

中国药科大学 学 报 470 Jo rnal of China Pharmace tical University 2006, 37 (5) :470- 473 X 1 2 1 徐建平 , 方慧生, 相秉仁 1 2 ( ; , 210009) 本文用常用的得分矩阵代替传统的Qian 编码作为神经网络的输入层预测了200 个蛋白质 级结构结果 表明: 以常用得分矩阵作为输入层的预测结果要优于Qian 编码的预测性能在200 个蛋白质中, 共有9 个蛋白质的预测精 度达到目前国际先进水平, 即80% 这说明该方法具有 一定的可行性 神经网络; 得分矩阵; 蛋白质 级结构预测 Q518. 2 A 1000- 5048(2006) 05- 0470- 04 Prediction of the protein secondary struct ure with common score matrix based neural network 1 2 1 XU Jian-ping , FANG H i-sheng ,XIANG Bing-ren 1Centerf or InstrumentAnalysis; 2Department of Bioinf ormatics, School of Lfi e Science Technology , China Pharmaceutical niversity, Nanj ing 210009, China Abstract The present paper describes the artificial ne ral network for the prediction of the protein secondary str ct re on the basis of common score matrix instead of Qian code as the inp t layer. Based on the predicted secondary str c- t re of 200 proteins, it was fo nd that the performance of the score matrix was a little better than that of Qian code. A- mong these 200 proteins, the predicted precision of 9 proteins was s perior to 80% , the well-recognized pper limit in the field of predicting the protein secondary str ct re. Also,there were no significant difference among res lts basedon a variety of score matrices1 Key words artificial ne ral network; score matrix; prediction of protein second str ct re 60 , 1 , Ala: ( 1, 0, 0, 0, ,, 0) ; Arg: ( 0, 1, 0, ,, 0) [ 1] [2] [3] Cho-Fasman Garnier [4] [5, 6] , (