基于卷积神经网络的细菌转录终止子预测  

Prediction of bacterial transcriptional terminators by using convolutional neural network

作　　者：金冬 张萌 贾藏芝 JIN Dong;ZHANG Meng;JIA Cangzhi(School of Science,Dalian Maritime University,Dalian 116026,Liaoning,China)

机构地区：[1]大连海事大学理学院,辽宁大连116026

出　　处：《生物信息学》2022年第3期182-188,共7页Chinese Journal of Bioinformatics

基　　金：国家自然科学基金(No.62071079).

摘　　要：在遗传学中,终止子是位于poly(A)位点下游、长度在数百碱基以内、包含多个回文序列、具有终止转录功能的DNA结构域,其主要作用是使转录终止。在原核生物基因组中有两类转录终止子,即Rho⁃dependent因子和Rho⁃independent因子。在本项研究中,提出了一种新的预测模型(TermCNN)来快速准确地识别细菌转录终止子。该模型将具有代表性的6⁃mer特征子集(2537个特征)和电子—离子相互作用伪电位(EIIP)作为输入向量,利用卷积神经网络(CNN)构建预测模型。五折交叉验证和独立测试的结果表明该模型优于最新的预测模型iTerm⁃PseKNC。值得注意的是,该模型在跨物种试验中具有明显的优势。它可以高度精确地预测大肠杆菌(E.coli)和枯草芽孢杆菌(B.subtilis)的转录终止子。In genetics,a transcriptional terminator is a DNA domain located downstream of poly(A)site within a length of hundreds of bases,which contains multiple palindrome sequences and has the function of terminating transcription.Two classes of transcriptional terminators,Rho⁃dependent and Rho⁃independent have been found in prokaryotic genomes.In this study,a novel model(Term CNN)was proposed for identifying bacterial transcriptional terminators rapidly and accurately.The model combined representative 6⁃mer sub⁃set(2537 features)and electron⁃ion interaction pseudopotentials(EIIP)of nucleotides as input parameters,and convolutional neural network(CNN)was utilized to train and optimize the model.Extensive 5⁃fold cross⁃validation and independent tests showed that the model outperformed the latest prediction model iTerm⁃PseKNC.It is especially noted that the model achieved obviously superiority on cross⁃species tests.In summary,the proposed model can predict transcriptional terminators of Escherichia coli(E.coli)and Bacillus subtilis(B.subtilis)with high accuray.

关 键 词：转录终止子 深度学习 特征选择 卷积神经网络 

分 类 号：Q939.1[生物学—微生物学]