基于深度最大输出网络的抗体种型特异性B细胞表位预测（英文）  

Deep Maxout Networks Applied to Antibody Class-Specific B-Cell Epitopes Prediction

作　　者：练瑶[1] 黄泽炽[2] 葛猛[3] 

机构地区：[1]清华大学生命科学学院,北京100084 [2]中国科学院生物物理研究所,北京100101 [3]中国科学院基因组科学及信息重点实验室,北京100101

出　　处：《激光生物学报》2016年第1期15-19,共5页Acta Laser Biology Sinica

基　　金：973 Program(2009CB918801)

摘　　要：B细胞表位研究有助于肽段疫苗研制,抗体研制以及疾病诊断和治疗研究。不同的B细胞表位诱导免疫系统产生不同的抗体种型,探索研究能够诱导特异性抗体产生的B细胞表位具有重要意义。基于二肽组成特征,利用深度最大输出网络算法训练构建三个二类分类器,分别对应诱导三种不同特异性抗体的B细胞表位,即Ig A表位,Ig E表位以及Ig G表位。通过五折交叉验证训练和测试这三个分类器,获得AUC的值分别为0.78,0.93以及0.78。Ig A表位和Ig E表位分类器的预测能力优于其它Ig A表位和Ig E表位分类器,Ig G表位分类器和其它Ig G表位分类器的预测能力相当。B-cell epitopes( BCEs) have been studied extensively due to their immunological applications,such as peptide-based vaccine development,antibody production,and disease diagnosis and therapy. The immune system produces different isotypes of antibodies against different BCEs in an antigen. It is critical to develop method for predicting and identifying antibody class-specific BCEs in an antigen. Based on dipeptide composition feature,three binary classifiers have been developed to identify and predict Ig A BCEs,Ig E BCEs and Ig G BCEs respectively using deep maxout network( DMN). A 5-fold cross-validation test was performed to evaluate the performance. Areas under the receiver operating characteristic curve( AUCs) were 0. 78,0. 93 and 0. 78 for Ig A,Ig E and Ig G BCEs predictors respectively. Comparison results demonstrate that Ig A and Ig E BCEs predictors outperform other Ig A and Ig E BCEs predictors; Ig G BCEs predictor shows comparable ability to other Ig G BCEs predictors.

关 键 词：深度最大输出网络 抗体种型 B细胞表位 预测 

分 类 号：Q6[生物学—生物物理学]