论文部分内容阅读
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 contribute to peptide vaccine development, antibody development and disease diagnosis and treatment research. Different B cell epitopes induce the immune system to produce different antibody types, exploring the importance of studying B cell epitopes that can induce the production of specific antibodies. Based on the characteristics of dipeptide composition, three class II classifiers were trained by using the maximum output network algorithm to induce B cell epitopes corresponding to three different specific antibodies, namely Ig A epitope, Ig E epitope and Ig G epitope Bit. Training and testing of the three classifiers through five-fold cross-validation resulted in AUC values of 0.78, 0.93 and 0.78, respectively. The predictive power of Ig A epitopes and Ig E epitope classifiers is superior to other Ig A epitopes and Ig E epitope classifiers, with comparable predictive power for Ig G epitope classifiers and other Ig G epitope classifiers.