北京生物医学工程

基于三维卷积神经网络的RNA结构上镁离子结合位点预测

Prediction of magnesium binding sites in RNA structures based onthree-dimensional convolutional neural network

作者：赵彦彭巩卫康刘洋王京京李春华

单位：北京工业大学环境与生命学部(北京100124) 通信作者:李春华。E-mail: chunhuali@ bjut. edu. cn <p> </p>

关键词： RNA-Mg2+结合位点;三维卷积神径网络;RNA三维结构;特征贡献;局部坐标系

分类号：R318 <p> </p>

出版年·卷·期（页码）：2021·40·6（570-576）

摘要：

目的核糖核酸(ribonucleic acid,RNA)在许多生命过程中起着关键作用，它几乎参与了遗传信息转录和翻译的各个方面。镁离子(M/+)是RNA折叠成稳定三级结构所必需的，而且常常与核酶的催化活性有关。尽管实验解析的RNA结构越来越多，但实验确定RNA结构中的离子存在诸多困难。三维卷积神经网络(three-dimensional convolutional neural network,3D-CNN)能够直接从原始数据中学习有用的特征，已被应用于一些基于结构的预测工作。为此本文拟提出一种基于3D-CNN的RNA结构上 Mg?+结合位点的预测方法RNAMgo方法首先收集蛋白质数据库(protein data bank,PDB)中的RNA结构构建了高分辨率的非冗余RNA-M/+结合位点数据库(113个结构)；随后，对每个结合位点建立局部坐标系，以周围微环境(碳、氧、氮、磷原子和电荷)构建5个特征通道，将RNA三级结构作为3D图像送入3D-CNN模型预测RNA-Mg2+结合位点;最后，使用五折交叉验证评估模型的性能。结果独立测试集上的结果表明，在识别RNA结构中的Mg，+位点上,RNAMg优于目前最先进的方法。结论RNAMg可以识别出RNA结构中的M/+结合位点,对理解RNA折叠、核酶催化反应等各种关键生物学过程及相关疾病的产生机制有重要意义。

Objective Ribonucleic acids ( RNA) play an important role in many life processes. They take part in almost all the aspects of gene transcription and translation. Magnesium ions ( Mg2+ ) are essential for RNA folding into stable tertiary structures and are often involved in the catalytic activity of ribozymes. Despite many RNA structures have been resolved experimentally, accurately detecting them still faces many challenges experimentally. Three-dimensional convolutional neural network ( 3D - CNN) are capable of learning useful

features directly from raw data, and have been used in many structure-based prediction works. Therefore, this paper proposes a 3D - CNN based method for predicting Mg2+ binding sites in RNA structure, RNAMg. Methods Firstly, a database of RNA-Mg2+ binding sites (113 RNAs) is constructed by collecting RNA structures from PDB (protein data bank) database. Then, a local coordinate system is established for each binding site, the ambient microenvironment is represented as five ' channels' ( corresponding to atom types of carbon, oxygen, nitrogen and phosphorus and charge) ,and the tertiary structure of RNA is sent into the 3D-CNN model as a 3D image to predict the RNA-Mg2+ binding sites. In the end, the performance of the model is evaluated by 5-fold cross validation. Results Results on an independent test set show that RNAMg is superior to state-of-the-art methods in identifying Mg2+ binding sites in RNA structures. Conclusions RNAMg can identify the Mg2+ binding sites in RNA structures, which is of great significance in understanding various key biological processes such as RNA folding, ribozyme-catalyzed reactions and the generation mechanism of related diseases.

参考文献：

[1 ] Doudna JA, Cech TR. The chemical repertoire of natural ribozymes [J]. Nature, 2002,418 (6894) ：222-228.

[2 ] Morris KV, Mattick JS. The rise of regulatory RNA [J]. Nature Reviews. Genetics,2014,15(6) ：423—437.

[3 ] Bowman JC, Lenz TK, Hud NV, et al. Cations in charge: magnesium ions in RNA folding and catalysis [ J ]. Current Opinion in Structural Biology, 2012,22(3) ：?262-272.

[4 ] Wang J, Xiao Y. Types and concentrations of metal ions affect local structure and dynamics of RNA [ J]. Physical Review E, 2016,94(4-1) ：040401.

[5 ] Yamauchi T, Miyoshi D, Kubodera T, et al. Roles of Mg2+?in TPP—dependent riboswitch [ J ]. FEBS Letters, 2005,579 ( 12)：?2583-2588.

[6 ] Schnabl J, Sigel RKO. Controlling ribozyme activity by metal ions [J ]. Current Opinion in Chemical Biology, 2010, 14 ( 2)：?269-275.

[7 ] Herschlag D, Cech TR. Catalysis of RNA cleavage by the tetrahymena-thermophila ribozyme . 1. Kinetic description of the reaction of an RNA substrate complementary to the active-site [J]. Biochemistry,1990,29(44) ：?10159-10171.

[8 ] Stellos K, Gatsiou A, Stamatelopoulos K, et al. Adenosine-to- inosine RNA editing controls cathepsin S expression in atherosclerosis by enabling HuR-mediated post-transcriptional regulation[ J] . Nature Medicine,2016,22( 10) ：?1140—1150.

[9 ] Cruz-Le6n S, Schwierz N. Hofmeister series for metal-cation-RNA interactions: the interplay of binding affinity and exchange kinetics [J]. Langmuir, 2020,36 (21) ：?5979-5989.

[10] Hu X,Dong Q,Yang J,et al. Recognizing metal and acid radical

ion-binding sites by integrating ab initio modeling with templatebased transferals[ J]. Bioinfbrmatics,2016,32(21) :3260-3269.

[11 ] Sfinchez-Aparicio JE, Tiessler-Sala L, Velasco-Cameros L, et al. BioMetAU: identifying metal-binding sites in proteins from backbone preorganization [ J ]. Journal of Chemical Information and Modeling,2021,61(l) ：311-323.

[12]?Philips A, Milanowska K, Lach G, et al. MetalionRNA: computational predictor of metal-binding sites in RNA structures [J]. Bioinformatics,2012,28(2) : 198—205.

[13]?Wang K, Jian Y, Wang H, et al. RBind: computational network method to predict RNA binding sites[ J]. Bioinfbrmatics,2018, 34(18) ：3131-3136.

[14]?Su H, Peng Z, Yang J. Recognition of small molecule-RNA binding sites using RNA sequence and structure [ J ]. Bioinfbrmatics,2021,37(1) :36-42.

[15]?Li J,Zhu W,Wang J,et al. RNA3DCNN：local and global quality assessments of RNA 3D structures using 3D deep convolutional neural networks [ J ]. PLoS Computational Biology, 2018, 14 (11):1-18.

[16]?Kozlovskii I, Popov P. Spatiotemporal identification of druggable binding sites using deep learning[ J] . Communications Biology, 2020,3(1)：618.

[17]?Berman HM, Westbrook J, Feng Z, et aL The protein data bank [J]. Nucleic Acids Research,2000,28(1):235-242.

[18]?Gong S, Zhang C, Zhang Y. RNA-align: quick and accurate alignment of RNA 3D structures based on size-independent TM- scoreRNA [ J ]. Bioinformatics, 2019,35(21) : 4459—4461.

[19]?Chapman EG,Costantino DA,Rabe JL,et al. The structural basis of pathogenic subgenomic flavivirus RNA ( sfRNA) production [J]. Science,2014,344(6181) ：307-310.

[20]?Correll CC,Munishkin A,Chan YL,et al. Crystal structure of the ribosomal RNA domain essential for binding elongation factors [J]. Proceedings of the National Academy of Sciences of the United States of America, 1998,95(23) ：?13436-13441.

[21 ] Correll CC, Freeborn B, Moore PB, et al. Metals, motifs, and recognition in the crystal structure of a 5S rRNA domain [ J ]. CeU,1997,91(5) ：705-712.

[22] Bonneau E, Legault P, NMR localization of divalent cations at the active site of the neurospora?VS ribozyme provides insights into RNA -metal-ion interactions [ J]. Biochemistry, 2014, 53 ( 3 )；?579-590.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录