基于深度学习预测赖氨酸巴豆酰化位点

Prediction of Lysine Bastoylation Sites Based on Deep Learning Method

巴豆酰化是一种新发现的蛋白翻译后修饰,参与细胞调控和人类疾病的发生。传统的蛋白翻译后修饰方法往往费时费力,因此建立有效的预测器是非常有必要的。本研究提出一种新的蛋白质翻译后修饰预测方法Cro-Deep。首先,利用二元编码(BE)、增强氨基酸组成(EAAC)、BLOSUM62转化为数字信息并进行融合;其次,使用GRU分类器对巴豆酰化位点进行预测;最后,利用十折交叉验证对模型进行检验。结果表明:训练集的ACC、MCC、和AUC值达到87.16%,0.743 7和0.935 7,独立测试集ACC、MCC、和AUC值达到91.54%, 0.831 3和0.961 5。实验结果表明,本研究提出的Cro-Deep方法能够有效的鉴定巴豆酰化位点,提高蛋白质翻译后修饰的预测效果。

Crobitoylation is a newly discovered post-translational modification of proteins involved in cell regulation and human disease. Traditional protein post-translational modification methods are often time-consuming and laborious, so it is necessary to establish effective predictors. In this paper, a novel predictive method for post-translational modification of proteins, Cro-Deep, is proposed. Firstly, binary coding(BE), enhanced amino acid composition(EAAC) and BLOSUM62 are used to transform digital information and fuse. Secondly, CROton acylation sites are predicted by GRU classifier. The results show that the ACC, MCC, and AUC values of the training set reach 87.16%, 0.743 7 and 0.935 7, respectively, and ACC, MCC, and AUC values of the independent test set reach 91.54%, 0.831 3, and 0.961 5, respectively. The experimental results show that the Cro-Deep method proposed in this paper can effectively identify croton acylation sites and improve the prediction effect of protein post-translational modification.