基于卷积神经网络的基因剪接位点预测  

Prediction of Gene Splicing Sites Based on Convolution Neural Network

作　　者：李国斌[1] 杜秀全 李新路[1] 吴志泽[1] LI Guobin;DU Xiuquan;LI Xinlu;WU Zhize(School of Artificial Intelligence and Big Data,Hefei University,Hefei Anhui 230601,China;Key Laboratory of Intelligent Computing and Signal Processing,Ministry of Education,Anhui University,Hefei Anhui 230601,China;School of Computer Science and Technology,Anhui University,Hefei Anhui 230601,China)

机构地区：[1]合肥学院人工智能与大数据学院,安徽合肥230601 [2]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230601 [3]安徽大学计算机科学与技术学院,安徽合肥230601

出　　处：《盐城工学院学报（自然科学版）》2020年第2期20-24,共5页Journal of Yancheng Institute of Technology:Natural Science Edition

基　　金：安徽省自然科学基金青年项目(1908085QF);安徽高校自然科学研究重点项目(KJ2019A0835);合肥学院科研发展项目基金(18ZR19ZDA)。

摘　　要：研究剪接位点可以更深入地探索剪接机制和基因预测方法,准确预测剪接位点至关重要。基于深度学习技术提出一种新的预测方法,无需人工提取样本特征,以基因序列的K-MER编码向量作为输入,采用训练后的卷积神经网络(CNN)模型进行预测。基于人类基因HS3D供体数据集,与传统机器学习方法进行预测比较,结果表明预测模型的主要性能指标,包含马修斯相关系数(MCC)、灵敏度(SN)均超过传统的机器学习方法。The study of splicing sites can further explore splicing mechanisms and gene prediction methods.It is very important to predict splice sites accurately.Based on the deep learning technology,a new prediction method is proposed.This method does not need to extract sample features manually.The K-MER coding vector of gene sequence is used as input,and the convolutional neural network(CNN)model after training is used for prediction.Based on the human gene HS3D donor data set,this model was compared with the traditional machine learning methods for prediction.The results showed that the main performance indicators of the prediction model,including Matthews correlation coefficient(MCC)and sensitivity(SN),exceeded the traditional machine learning methods.

关 键 词：深度学习 卷积神经网络 剪接位点预测 K-MER编码 

分 类 号：TP183.1[自动化与计算机技术—控制理论与控制工程]