基于双向长短时记忆网络和注意力机制的RNA m5C甲基化位点预测  

Prediction of RNA m5C Methylation Sites Based on Bi-directional Long Short-term Memory and Attention Mechanism

作　　者：胡梦 李慧敏[1] 唐轶[1] 王煜 陈鹏辉 HU Meng;LI Hui-Min;TANG Yi;WANG Yu;CHEN Peng-Hui(School of Mathematics and Computer Science,Yunnan Minzu University,Kunming 650504,China)

机构地区：[1]云南民族大学数学与计算机科学学院,昆明650504

出　　处：《中国生物化学与分子生物学报》2023年第2期303-310,共8页Chinese Journal of Biochemistry and Molecular Biology

基　　金：国家自然科学基金项目(No.61866040);云南民族大学研究生科研项目(No.SJXY-2021-011)资助。

摘　　要：RNA 5-甲基胞嘧啶(m5C)修饰在许多生物过程中发挥重要的作用,对m5C位点的准确识别有助于更好地理解其生物学功能,所以识别m5C甲基化位点十分必要。尽管已发展了多种识别m5C甲基化位点的机器学习方法,但预测能力仍有待提高。本文基于双向长短时记忆网络和注意力机制,提出了一种预测RNA m5C甲基化位点的深度学习算法。用该方法在人、小鼠、酿酒酵母和拟南芥共4种生物的RNA m5C数据集上进行实验,m5C位点预测AUC值分别达到92.5%、99.7%、93.6%和86.5%。与现有预测方法相比,该方法具有较好的预测性能,并且具有更优的泛化能力,为RNA m5C甲基化位点预测提供了一种新方法。RNA 5-methylcytosine(m5C) modification plays an important role in many biological processes,and accurate identification of m5C sites helps to better understand their biological function,so it is necessary to identify the methylation sites of m5C.Although several machine learning methods have been developed to identify the methylation sites of m5C,the ability of prediction remains to be improved.In this paper,we proposed a deep learning algorithm,to predict the methylation site of m5C based on bidirectional short and long-term memory network and attention mechanism.The predicted AUC of m5C was 92.5%,99.7%,93.6% and 86.5% for Homo sapiens,Mus musculus,Saccharomyces cerevisiae and Arabidopsis thaliana RNA m5C datasets respectively.Compared with the existing prediction methods,this method has better prediction performance and better generalization ability,which provides a new method for RNA m5C methylation site prediction.

关 键 词：双向长短时记忆网络 注意力机制 m5C甲基化位点 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] Q52[自动化与计算机技术—计算机科学与技术]