摘要
N7-甲基鸟苷(N7-methylguanosine,m7G)修饰在RNA修饰中普遍存在,识别m7G位点对认识m7G功能和深入了解人类疾病具有重要意义.目前关于m7G位点的识别方法大多基于传统机器学习,需要手动输入筛选最优特征,存在特征冗余问题.为了解决以上问题,提出一种多维度卷积神经网络,该方法基于卷积神经网络构建,并在卷积的基础上增加空间空洞卷积层,采用空洞空间卷积池化金字塔模块获得多尺度序列信息特征,以扩大模型的感受野,使得提取的特征更加全面.基于相同的m7G位点序列数据,将多维度卷积神经网络模型的m7G位点预测能力与几种已有算法进行比较,结果表明,多维度卷积神经网络模型的预测性能优于现有算法.
N7-methylguanosine(m7G)modification is ubiquitous in RNA modification,and the identification of m7G sites is of great significance for understanding the function of m7G and gaining insights into human diseases.At present,most recognition methods for m7G sites are based on traditional machine learning,which requires manual input to screen the optimal features,resulting in feature redundancy.In order to solve the above problems,a multi-dimensional convolutional neural network was proposed.Based on the construction of convolutional neural network,a spatial void convolution layer was added on the basis of convolution,and a spatial pyramid pooling module was used to obtain multi-scale sequence information features,so as to enlarge the model s receptive field and make the extracted features more comprehensive.Based on the same m7G site sequence data,the m7G site prediction ability of multi-dimensional convolutional neural network model was compared with several existing algorithms,and the results show that the prediction performance of multi-dimensional convolutional neural network model is better than the existing algorithms.
作者
王煜
李慧敏
唐轶
胡梦
陈鹏辉
WANG Yu;LI Hui-min;TANG Yi;HU Meng;CHEN Peng-hui(School of Mathematics and Computer Science,Yunnan Minzu University,Kunming 650500,China)
出处
《云南民族大学学报（自然科学版）》
CAS
2024年第6期753-759,785,共8页
Journal of Yunnan Minzu University:Natural Sciences Edition
基金
国家自然科学基金(61866040)
云南省研究生优质课程建设项目(云学位[2022]8号)
云南民族大学数学与计算机科学学院研究生科研项目(SJXY-2021-015)。
关键词
多维度卷积神经网络
空洞卷积
m7G甲基化
深度学习
multi-dimensional convolutional neural network
atrous convolution
N7-methylguanosine
deep learning
