摘要
本文针对传统蛋白质相互作用预测模型预测精度不够高的问题,提出一种改进的深度玻尔兹曼机(DBM)模型以更精确地预测蛋白质的相互作用。首先,将多尺度特征组提取和自协方差编码方法结合编码序列特征,并利用DBM自动筛选有效特征。同时,为了避免采用sigmoid或tanh激活函数在深度网络中出现过饱和的问题,本文采用Re LU改进的深度玻尔兹曼机(RBM),使网络具备稀疏性,从而避免模型过拟合,加快收敛速度。在酵母菌PPIs数据集上,本文算法达到了92.27%的准确率,优于传统的方法。
To address the problem that prediction accuracy is not high enough in traditional protein interaction prediction model, we present a new improved deep boltzmann machine model(DBM) to predict protein interactions accurately. Firstly, combining a novel Multi-scale Continuous and Discontinuous(MCD) feature representation and autocovariance approach to encoding protein sequence, and employing DBM to automatically select an effective feature. In addition, in order to avoid saturation by using sigmoid or tanh activation function in depth network, Re LU modified restricted boltzmann machine(RBM) is selected to improve the sparsity of the network, to avoid over-fitting,and to improve the convergence rate. The prediction accuracy of proposed model on yeast PPIs data set achieved92.27%, which indicates that our performance better than the previous related models.
出处
《基因组学与应用生物学》
CAS
CSCD
北大核心
2015年第8期1791-1797,共7页
Genomics and Applied Biology
基金
中央高校基本科研业务费专项资金资助
