基于CNN和Bi-LSTM模型的蛋白质甲基化位点识别

打开文本图片集

关键词：甲基化；卷积神经网络；双向长短时记忆网络；特征融合；位点识别DOI：10.15938/j. jhust.2025.02.009中图分类号：TP391.4 文献标志码：A 文章编号：1007-2683（2025）02-0082-09

Abstract：MethylationisaproteinPost-Translational Modification（PTM）thatregulatescellfunction，whichcanprovideguidance andhelpforresearchinthefieldsofgeneregulationanddiseaseprediction.Atpresent，therearesomeproblemsintheresearchof methylationsiterecogitin，suchasfewlabeleddtasets，nsuffcentpositivesmpledataandlowrecogitionaccracyofhylation research.Inordertosolvetheseproblems，thispaperproposesaproteinmethylationsiterecognitionmethodbasedonConvolutional NeuralNetwork（CNN）andBi-directionalLongShor-TermMemory（Bi-LSTM）model.Ourmodelisdividedintotwobranches.The CNNbranchusesadenseconnectionmethodtomakethefeatureinformationofachlayerfull transmitedandshared.ThestackedBiLSTMbranchesobtainbidirectionallong-termdependenciesinthesequence，andthenthetwobranchesperform featurefusionfor methylationrecognition.ExperimentsshowthattheAccuracy（ACC），F1Score（F1score）andMatthewsCorrelationCoficient （MCC）obtainedbyusingourmodeltoidentifymethylationsitesareO.8519，0.8494andO.7284，respectively.Compared with other methods，the model has better performance.

Keywords;methylation；convolutionalneuralnetwork；bi-directionallongshrt-termmemory；featurefusion；siterecogition

0 引言

甲基化是一种翻译后修饰（post-translationalmodification，PTM），通过向蛋白质结构中添加甲基来修饰蛋白质的功能和构象，在表观遗传过程中发挥着重要的作用[]。（剩余15029字）