摘要
考虑到现有的基于序列的蛋白质相互作用预测方法均采用单一的特征提取方法,具有一定的局限性,提出一种方法。用元学习策略作为分类器融合策略,并集成多种蛋白质序列特征提取方法。在10 702对酿酒酵母蛋白质对数据集上,得到97.28%的预测精度,优于目前现有方法的平均水平,在独立测试集上同样具有优秀的表现,实验结果表明,该方法有效提高了蛋白质相互作用预测的准确率。
Considering that the existing methods for protein-protein interactions (PPIs) based on protein sequence use the singlefeature extraction and have certain limitations ’ a method based on multiple feature extraction for protein sequence was proposed. Experiments were carried out on the data set with 10 702 Saccharomyces cerevisiae protein pairs. tion accuracy of the proposed method reach 97. 28% ’ which is superior to the average of dent tett set’ the proposed method also shows excellent performance’ indicating that it effectively improves the accuracy of pre-diction of PPIs.
出处
《计算机工程与设计》
北大核心
2018年第1期86-89,254,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61273225
61502356)