Volume 50 Issue 5
Sep.  2021
Article Contents

SU Wei, SUN Zijie, YUE Peng, LIN Hao. A Brief Review for Identifying Prokaryotic Promoters Based on Computational Biology[J]. Journal of University of Electronic Science and Technology of China, 2021, 50(5): 667-675. doi: 10.12178/1001-0548.2021201
Citation: SU Wei, SUN Zijie, YUE Peng, LIN Hao. A Brief Review for Identifying Prokaryotic Promoters Based on Computational Biology[J]. Journal of University of Electronic Science and Technology of China, 2021, 50(5): 667-675. doi: 10.12178/1001-0548.2021201

A Brief Review for Identifying Prokaryotic Promoters Based on Computational Biology

doi: 10.12178/1001-0548.2021201
  • Received Date: 2021-07-29
  • Rev Recd Date: 2021-08-30
  • Available Online: 2021-09-28
  • Publish Date: 2021-09-28
  • As a key region of deoxyribonucleic acid (DNA), prokaryotic promoter contains the conserved sequence required for specific binding of ribonucleic acid (RNA) polymerase and transcription initiation, and plays an important role in transcription regulation. However, due to the limitations of experimental methods that are long experimental period and high cost, the identification of prokaryotic promoter sequences remains a major challenge. With the development of computer technology, dozens of prokaryotic promoter identification methods based on computational biology have emerged, which show a high degree of diversity in terms of data quality, dataset size, extracted features, feature selection techniques, classification algorithms and evaluation strategies. Thus, there is an urgent need to systematically compare and summarize these methods so as to improve and further develop prokaryotic promoter recognition techniques.
  • [1] PERRON G G, WHYTE L, TURNBAUGH P J, et al. Functional characterization of bacteria isolated from ancient arctic soil exposes diverse resistance mechanisms to modern antibiotics[J]. Plos One, 2015, 10(3): e0069533. doi:  10.1371/journal.pone.0069533
    [2] COOK H, USSERY D W. Sigma factors in a thousand E. coli genomes[J]. Environ Microbiol, 2013, 15(12): 3121-3129. doi:  10.1111/1462-2920.12236
    [3] BARRIOS H, VALDERRAMA B, MORETT E. Compilation and analysis of sigma(54)-dependent promoter sequences[J]. Nucleic Acids Res, 1999, 27(22): 4305-4313. doi:  10.1093/nar/27.22.4305
    [4] JANGA S C, COLLADO-VIDES J. Structure and evolution of gene regulatory networks in microbial genomes[J]. Res Microbiol, 2007, 158(10): 787-794. doi:  10.1016/j.resmic.2007.09.001
    [5] LEE S J, GRALLA J D. Sigma38 (rpoS) RNA polymerase promoter engagement via -10 region nucleotides[J]. J Biol Chem, 2001, 276(32): 30064-30071. doi:  10.1074/jbc.M102886200
    [6] POTVIN E, SANSCHAGRIN F, LEVESQUE R C. Sigma factors in Pseudomonas aeruginosa[J]. Fems Microbiol Rev, 2008, 32(1): 38-55. doi:  10.1111/j.1574-6976.2007.00092.x
    [7] ABRIL A G, RAMA J L R, SANCHEZ-PEREZ A, et al. Prokaryotic sigma factors and their transcriptional counterparts in Archaea and Eukarya[J]. Appl Microbiol Biot, 2020, 104(10): 4289-4302. doi:  10.1007/s00253-020-10577-0
    [8] 丁辉, 邓恩泽, 陈伟, 等. 细菌σ54启动子序列分析与预测[J]. 电子科技大学学报, 2015, 44(1): 147-149. doi:  10.3969/j.issn.1001-0548.2015.01.025

    DING H, DENG E Z, CHEN W, et al. The sequence analysis and prediction of σ54 promoter in bacteria[J]. Journal of University of Electronic Science and Technology of China, 2015, 44(1): 147-149. doi:  10.3969/j.issn.1001-0548.2015.01.025
    [9] BURDEN S, LIN Y X, ZHANG R. Improving promoter prediction for the NNPP2.2 algorithm: A case study using Escherichia coli DNA sequences[J]. Bioinformatics, 2005, 21(5): 601-607. doi:  10.1093/bioinformatics/bti047
    [10] WANG H, BENHAM C J. Promoter prediction and annotation of microbial genomes based on DNA sequence and structural responses to superhelical stress[J]. Bmc Bioinformatics, 2006, 7: 248. doi:  10.1186/1471-2105-7-248
    [11] POLAT K, GUNES S. A novel approach to estimation of E-coli promoter gene sequences: Combining feature selection and least square support vector machine (FS_LSSVM)[J]. Appl Math Comput, 2007, 190(2): 1574-1582.
    [12] RANGANNAN V, BANSAL M. Identification and annotation of promoter regions in microbial genome sequences on the basis of DNA stability[J]. J Biosciences, 2007, 32(5): 851-862.
    [13] RANGANNAN V, BANSAL M. Relative stability of DNA as a generic criterion for promoter prediction: Whole genome annotation of microbial genomes with varying nucleotide base composition[J]. Mol Biosyst, 2009, 5(12): 1758-1769. doi:  10.1039/b906535k
    [14] BLAND C, NEWSOME A S, MARKOVETS A A. Promoter prediction in E-coli based on SIDD profiles and artificial neural networks[J]. Bmc Bioinformatics, 2010, 11: S17.
    [15] DE J A, PIETERSMA H, CORDES M, et al. PePPER: A webserver for prediction of prokaryote promoter elements and regulons[J]. Bmc Genomics, 2012, 13: 299. doi:  10.1186/1471-2164-13-299
    [16] DI S M, PINATEL E, TALA A, et al. G4PromFinder: An algorithm for predicting transcription promoters in GC-rich bacterial genomes based on AT-rich elements and G-quadruplex motifs[J]. Bmc Bioinformatics, 2018, 19(1): 36. doi:  10.1186/s12859-018-2049-x
    [17] PEREZ-BELLO A, MUNTEANU C R, UBEIRA F M, et al. Alignment-free prediction of mycobacterial DNA promoters based on pseudo-folding lattice network or star-graph topological indices[J]. J Theor Biol, 2009, 256(3): 458-466. doi:  10.1016/j.jtbi.2008.09.035
    [18] GORDON J J, TOWSEY M W, HOGAN J M, et al. Improved prediction of bacterial transcription start sites[J]. Bioinformatics, 2006, 22(2): 142-148. doi:  10.1093/bioinformatics/bti771
    [19] TOWSEY M, TIMMS P, HOGAN J, et al. The cross-species prediction of bacterial promoters using a support vector machine[J]. Comput Biol Chem, 2008, 32(5): 359-366. doi:  10.1016/j.compbiolchem.2008.07.009
    [20] RANI T S, BHAVANI S D, BAPI R S. Analysis of E. coli promoter recognition problem in dinucleotide feature space[J]. Bioinformatics, 2007, 23(5): 582-588. doi:  10.1093/bioinformatics/btl670
    [21] LI Q Z, LIN H. The recognition and prediction of sigma70 promoters in Escherichia coli K-12[J]. J Theor Biol, 2006, 242(1): 135-141. doi:  10.1016/j.jtbi.2006.02.007
    [22] LIN H, LI Q Z. Eukaryotic and prokaryotic promoter prediction using hybrid approach[J]. Theory Biosci, 2011, 130(2): 91-100. doi:  10.1007/s12064-010-0114-8
    [23] HE W, JIA C, DUAN Y, et al. 70ProPred: A predictor for discovering sigma70 promoters based on combining multiple features[J]. BMC Syst Biol, 2018, 12(Suppl 4): 44.
    [24] LAI H Y, ZHANG Z Y, SU Z D, et al. iProEP: A computational predictor for predicting promoter[J]. Mol Ther-Nucl Acids, 2019, 17: 337-346. doi:  10.1016/j.omtn.2019.05.028
    [25] WU Q Q, WANG J J, YAN H. An improved position weight matrix method based on an entropy measure for the recognition of prokaryotic promoters[J]. Int J Data Min Bioin, 2011, 5(1): 22-37. doi:  10.1504/IJDMB.2011.038575
    [26] DE A E, ECHEVERRIGARAY S, GERHARDT G J. BacPP: Bacterial promoter prediction—a tool for accurate sigma-factor specific assignment in enterobacteria[J]. J Theor Biol, 2011, 287: 92-99. doi:  10.1016/j.jtbi.2011.07.017
    [27] SONG K. Recognition of prokaryotic promoters based on a novel variable-window Z-curve method[J]. Nucleic Acids Res, 2012, 40(3): 963-971. doi:  10.1093/nar/gkr795
    [28] SILVA S A, FORTE F, SARTOR I T, et al. DNA duplex stability as discriminative characteristic for Escherichia coli sigma(54)- and sigma(28)- dependent promoter sequences[J]. Biologicals, 2014, 42(1): 22-28. doi:  10.1016/j.biologicals.2013.10.001
    [29] LIN H, DENG E Z, DING H, et al. iPro54-PseKNC: A sequence-based predictor for identifying sigma-54 promoters in prokaryote with pseudo k-tuple nucleotide composition[J]. Nucleic Acids Res, 2014, 42(21): 12961-12972. doi:  10.1093/nar/gku1019
    [30] LIU B, HAN L, LIU X, et al. Computational prediction of sigma-54 promoters in bacterial genomes by integrating motif finding and machine learning strategies[J]. IEEE/ACM Trans Comput Biol Bioinform, 2019, 16(4): 1211-1218. doi:  10.1109/TCBB.2018.2816032
    [31] ABBAS M M, MOHIE-ELDIN M M, EL-MANZALAWY Y. Assessing the effects of data selection and representation on the development of reliable E. coli sigma 70 promoter region predictors[J]. Plos One, 2015, 10(3): e0119721. doi:  10.1371/journal.pone.0119721
    [32] SHAHMURADOV I A, MOHAMAD RAZALI R, BOUGOUFFA S, et al. bTSSfinder: A novel tool for the prediction of promoters in cyanobacteria and Escherichia coli[J]. Bioinformatics, 2017, 33(3): 334-340.
    [33] LIN H, LIANG Z Y, TANG H, et al. Identifying sigma70 promoters with novel pseudo nucleotide composition[J]. IEEE ACM T Comput Bi, 2019, 16(4): 1316-1321.
    [34] RAHMAN M S, AKTAR U, JANI M R, et al. iPromoter-FSEn: Identification of bacterial sigma(70) promoter sequences using feature subspace based ensemble classifier[J]. Genomics, 2019, 111(5): 1160-1166. doi:  10.1016/j.ygeno.2018.07.011
    [35] RAHMAN M S, AKTAR U, JANI M R, et al. iPro70-FMWin: Identifying Sigma70 promoters using multiple windowing and minimal features[J]. Mol Genet Genomics, 2019, 294(1): 69-84. doi:  10.1007/s00438-018-1487-5
    [36] UMAROV R K, SOLOVYEV V V. Recognition of prokaryotic and eukaryotic promoters using convolutional deep learning neural networks[J]. Plos One, 2017, 12(2): e0171410. doi:  10.1371/journal.pone.0171410
    [37] WANG S, CHENG X, LI Y, et al. Image-based promoter prediction: A promoter prediction method based on evolutionarily generated patterns[J]. Sci Rep, 2018, 8(1): 17695. doi:  10.1038/s41598-018-36308-0
    [38] COPPENS L, LAVIGNE R. SAPPHIRE: A neural network based classifier for sigma70 promoter prediction in Pseudomonas[J]. Bmc Bioinformatics, 2020, 21(1): 415. doi:  10.1186/s12859-020-03730-z
    [39] LIU B, YANG F, HUANG D S, et al. iPromoter-2L: A two-layer predictor for identifying promoters and their types by multi-window-based PseKNC[J]. Bioinformatics, 2018, 34(1): 33-40. doi:  10.1093/bioinformatics/btx579
    [40] LIU B, LI K. iPromoter-2L2.0: Identifying promoters and their types by combining smoothing cutting window algorithm and sequence-based features[J]. Mol Ther Nucleic Acids, 2019, 18: 80-87. doi:  10.1016/j.omtn.2019.08.008
    [41] ZHANG M, LI F, MARQUEZ-LAGO T T, et al. MULTiPly: A novel multi-layer predictor for discovering general and specific types of promoters[J]. Bioinformatics, 2019, 35(17): 2957-2965. doi:  10.1093/bioinformatics/btz016
    [42] SHUJAAT M, WAHAB A, TAYARA H, et al. pcPromoter-CNN: A CNN-based prediction and classification of promoters[J]. Genes (Basel), 2020, 11(12): 1529. doi:  10.3390/genes11121529
    [43] AMIN R, RAHMAN C R, AHMED S, et al. iPromoter-BnCNN: A novel branched CNN-based predictor for identifying and classifying sigma promoters[J]. Bioinformatics, 2020, 36(19): 4869-4875. doi:  10.1093/bioinformatics/btaa609
    [44] LI F, CHEN J, GE Z, et al. Computational prediction and interpretation of both general and specific types of promoters in Escherichia coli by exploiting a stacked ensemble-learning framework[J]. Brief Bioinform, 2021, 22(2): 2126-2140. doi:  10.1093/bib/bbaa049
    [45] XIAO X, XU Z C, QIU W R, et al. IPSW(2L)-PseKNC: A two-layer predictor for identifying promoters and their strength by hybrid features via pseudo K-tuple nucleotide composition[J]. Genomics, 2019, 111(6): 1785-1793. doi:  10.1016/j.ygeno.2018.12.001
    [46] LE N Q K, YAPP E K Y, NAGASUNDARAM N, et al. Classifying promoters by interpreting the hidden information of DNA sequences via deep learning and combination of continuous fasttext N-grams[J]. Front Bioeng Biotechnol, 2019, 7: 305. doi:  10.3389/fbioe.2019.00305
    [47] TAYARA H, TAHIR M, CHONG K T. Identification of prokaryotic promoters and their strength by integrating heterogeneous features[J]. Genomics, 2020, 112(2): 1396-1403. doi:  10.1016/j.ygeno.2019.08.009
    [48] SANTOS-ZAVALETA A, SALGADO H, GAMA-CASTRO S, et al. RegulonDB v 10.5: Tackling challenges to unify classic and high throughput knowledge of gene regulation in E. coli K-12[J]. Nucleic Acids Research, 2019, 47(D1): D212-D220. doi:  10.1093/nar/gky1077
    [49] ISHII T, YOSHIDA K, TERAI G, et al. DBTBS: A database of Bacillus subtilis promoters and transcription factors[J]. Nucleic Acids Research, 2001, 29(1): 278-280. doi:  10.1093/nar/29.1.278
    [50] HUANG Y, NIU B F, GAO Y, et al. CD-HIT suite: A web server for clustering and comparing biological sequences[J]. Bioinformatics, 2010, 26(5): 680-682. doi:  10.1093/bioinformatics/btq003
    [51] SU W, LIU M L, YANG Y H, et al. PPD: A manually curated database for experimentally verified prokaryotic promoters[J]. Journal of Molecular Biology, 2021, 433(11): 166860.
    [52] CHEN W, LEI T Y, JIN D C, et al. PseKNC: A flexible web server for generating pseudo K-tuple nucleotide composition[J]. Anal Biochem, 2014, 456: 53-60. doi:  10.1016/j.ab.2014.04.001
    [53] PENG H C, LONG F H, DING C. Feature selection based on mutual information: Criteria of max-dependency, max-relevance, and min-redundancy[J]. IEEE T Pattern Anal, 2005, 27(8): 1226-1238. doi:  10.1109/TPAMI.2005.159
    [54] ZOU Q, ZENG J, CAO L, et al. A novel features ranking metric with application to scalable visual and bioinformatics data classification[J]. Neurocomputing, 2016, 173: 346-354. doi:  10.1016/j.neucom.2014.12.123
    [55] CHANG C C, LIN C J. LIBSVM: A library for support vector machines[J]. ACM Transactions on Intelligent Systems and Technology, 2011, 2(3): 1-27.
    [56] FISHER R A. The use of multiple measurements in taxonomic problems[J]. Annals of Human Genetics, 2012, 7(7): 179-188.
  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(2)  / Tables(2)

Article Metrics

Article views(5899) PDF downloads(104) Cited by()

Related
Proportional views

A Brief Review for Identifying Prokaryotic Promoters Based on Computational Biology

doi: 10.12178/1001-0548.2021201

Abstract: As a key region of deoxyribonucleic acid (DNA), prokaryotic promoter contains the conserved sequence required for specific binding of ribonucleic acid (RNA) polymerase and transcription initiation, and plays an important role in transcription regulation. However, due to the limitations of experimental methods that are long experimental period and high cost, the identification of prokaryotic promoter sequences remains a major challenge. With the development of computer technology, dozens of prokaryotic promoter identification methods based on computational biology have emerged, which show a high degree of diversity in terms of data quality, dataset size, extracted features, feature selection techniques, classification algorithms and evaluation strategies. Thus, there is an urgent need to systematically compare and summarize these methods so as to improve and further develop prokaryotic promoter recognition techniques.

SU Wei, SUN Zijie, YUE Peng, LIN Hao. A Brief Review for Identifying Prokaryotic Promoters Based on Computational Biology[J]. Journal of University of Electronic Science and Technology of China, 2021, 50(5): 667-675. doi: 10.12178/1001-0548.2021201
Citation: SU Wei, SUN Zijie, YUE Peng, LIN Hao. A Brief Review for Identifying Prokaryotic Promoters Based on Computational Biology[J]. Journal of University of Electronic Science and Technology of China, 2021, 50(5): 667-675. doi: 10.12178/1001-0548.2021201
  • 启动子通常位于基因上游,能与RNA聚合酶特异性结合并起始转录的一段DNA序列,作为转录起始过程的关键元件,激活RNA聚合酶与模板DNA结合,是基因表达和转录调节的起始步骤[1]

    原核生物RNA聚合酶中的σ因子可以特异性识别并结合启动子。在大肠杆菌中,存在多种σ因子,根据分子量可以分为7类,σ70、σ54、σ38、σ32、σ28、σ24、σ19,在已知的7类σ因子中前6类保守性极强,而σ19在大多数基因组中是缺失的[2]。每一类σ因子具有特定的生物学功能[3-6],σ70主要负责持家基因的转录;σ54被认为是参与氮代谢的调控因子以及控制一些辅助进程;σ38参与稳定期基因的调节;σ32是热休克σ因子(热激因子);σ28参与鞭毛的合成;σ24与极端热应激反应有关;σ19则参与对铁离子转运系统的调控。根据σ因子的同源性,可将其大致分为两类:一类是σ70家族,包括σ70、σ38、σ32、σ28、σ24、σ19;另一类是σ54家族。大肠杆菌基因组内的启动子类型依据与之结合的σ因子种类也可分为相应的类型。不同类型的启动子共有序列也有所差异。因此,启动子也依据被识别的片段分为σ70家族和σ54家族。如σ70启动子具有两个重要的基序区域,−10区和−35区,分别位于转录起始位点上游约10 bp和35 bp处。−10区含有保守序列“TATAAT”,又被称为Pribnow box或TATA box,富含腺嘌呤(adenine, A)和胸腺嘧啶(thymine, T),有助于DNA双链解螺旋分离;−35区则由6个保守的核苷酸“TTGACA”组成[7]。除了σ70因子,−10区和−35区也是被σ70家族其他因子识别的重要片段。相比之下,σ54启动子的共有序列及其位置与σ70启动子具有明显差异,在σ54启动子的−24区和−12区存在保守区域,其保守序列分别是“TGGCA[CT][GA]”和“TGC[AT][TA]”[8]

    启动子序列的鉴定对于研究基因表达、分析基因调控机制、研究基因结构以及注释基因信息至关重要。准确识别启动子的方法一般是依靠昂贵且耗时费力的实验检测方法,然而,在全基因组范围内进行检测是一项艰巨的任务。随着测序技术以及计算机技术的发展,越来越多生物的全基因组被测序出来,尤其是原核生物,因此出现了基于计算生物学的启动子预测方法,这些预测方法在不断地改进,有助于鉴别启动子序列。

    ToolsBenchmark dataset size (promoter)Sequence similarityFeature extraction/ selectionClassification algorithmEvaluation strategyAUC
    1.TLS-NNPP[9]771 (E.coli)/The empirical probability
    distribution of TSS-TLS distance
    ANNIndependent test/
    2.SIDD[10]500 (E.coli)/SIDDFLDIndependent test/
    3.FS_LSSVM[11]53 (E.coli)/A domain theory for promoters/
    C4.5 decision tree
    LSSVM10-fold cross-validation/
    4.Free energy[12]1044 (E.coli)
    879 (B.subtilis)
    /Free energyModified scoring functionIndependent test/
    5.PromPredict[13]1145 (E.coli) 615 (B.subtilis)
    82 (M.tuberculosis)
    /GC content; Average free energydifference between the average free energyTraining and validation/
    6.SIDD-ANN[14]1648 (E.coli)/SIDD profile dataANNIndependent test/
    7.PePPER[15]L.lactis/PWMHMM//
    8.G4PromFinder[16]3570 (S.coelicolor)
    2117 (P.aeruginosa)
    /AT-rich element and G-quadruplex motif-based algorithm/Independent test/
    9.LN-QSAR[17]135 (M.bovis)/Pseudo-folding 2D lattice graphLDAIndependent test/
    10.Ensemble-SVM[18]450 (E.coli σ70)/k-mer with location with respect to the TSS/ Symmetric uncertaintyEnsemble-SVM10-fold cross-validation/
    11.TSS-PREDICT[19]450 (E.coli σ70)
    205 (B.subtilis)
    26 (C.trachomatis)
    /Information Content; PWMEnsemble-SVMIndependent test/
    12.TSS-SLP[20]669 (E.coli σ70)/Dinucleotide Frequency FeaturesSLP5-fold cross-validation; Independent test/
    13.PCSF[21]683 (E.coli σ70)/Conversation of sequence segments; PCSFScore function10-fold cross-validation/
    14.IPMD[22]270 (B.subtilis σ43)
    741 (E.coli σ70)
    /PCSF; IDModified MD10-fold cross-validation0.847 (B.subtilis)
    0.920 (E.coli)
    15.70ProPred[23]741 (E.coli σ70)/PSTNPss; PseEIIPSVM5-fold cross-validation; Jackknife test0.990
    16.iProEP[24]270 (B.subtilis)
    741 (E.coli)
    ≤80%PseKNC; PCSF/ mRMR; IFSSVM10-fold cross-validation0.988 (B.subtilis)
    0.976 (E.coli)
    17.IPWM[25]683 (E.coli σ70)/Entropy-based conservative characteristics; Improved PWMScore function10-fold cross-validation/
    18.BacPP[26]1034 (E.coli)/Binary digitsANN(2,3,10)-fold cross-validation; Independent test/
    19.vw Z-curve[27]1401 (E.coli) 660 (B.subtilis)/variable-window Z-curve/ IFSPLS10-fold cross-validation/
    20.Stability[28]1035 (E.coli)/DNA duplex stabilityANN(2,3,10)-fold cross-validation/
    21.iPro54-PseKNC[29]161 (prokaryotic σ54)≤75%PseKNC/ F-score; IFSSVMJackknife test/
    22.Promote
    Predictor[30]
    161 (prokaryotic σ54)≤75%Motif profile-based ANF/ MRMDBagging; RF; SVM10-fold cross-validation; Independent test/
    23.meta-predictior[31]579 (E.coli σ70)≤45%sequence-based features; structure-based featuresMeta-predictorIndependent test0.850
    24.bTSSfinder[32]3597 (E.coli) 12797 (Nostoc) 351 (Synechocystis)
    1471 (S.elongatus)
    /PWM; Physicochemical properties/ Mahalanobis distanceANNIndependent test/
    25.iPro70-PseZNC[33]741 (E.coli σ70)/PseZNC/ F-score; IFSSVM5-fold cross-validation0.909
    26.iPromoter-FSEn[34]741 (E.coli σ70)/Nucleotide Statistics; k-mer; g-gapped k-mer; Approximate signal pattern count; Position specific occurences; Distribution of nucleotides/ Feature subspaceEnsemble learning10-fold cross-validation0.932
    27.iPro70-FMWin[35]741 (E.coli σ70)/k-mer; g-gapped k-mer; Pattern finding; Positioning distance count/ AdaboostLR10-fold cross-validation0.959
    28.CNNProm[36]839 (E.coli σ70)
    746 (B.subtilis)
    /one-hotCNN5-fold cross-validation/
    29.IBBP[37]1888 (E.coli σ70)/Image-based and evolutionary approachSVMIndependent test/
    30.SAPPHIRE[38]170 (P. aeruginosa and P. putida σ70)/one-hotANN5-fold cross-validation; Independent test/
    31.iPromoter-2L[39]2860 (E.coli)≤80%Multi-window-based PseKNCRF5-fold cross-validation; Jackknife test/
    32.iPromoter-2L2.0[40]2860 (E.coli)≤80%Smoothing Cutting Window algorithm; k-mer; PseKNCSVM; Ensemble learning5-fold cross-validation/
    33.MULTiPly[41]2860 (E.coli)≤80%Bi-profile bayes; KNN; k-mer;
    DAC/ F-score
    SVM5-fold cross-validation; Jackknife test; Independent test/
    34.pcPromoter-CNN[42]2860 (E.coli)≤80%one-hotCNN5-fold cross-validation; Independent test0.957
    35.iPromoter-BnCNN[43]2860 (E.coli)≤80%one-hot; k-mer; Structural
    properties
    CNN5-fold cross-validation; Independent test/
    36.SELECTOR[44]2860 (E.coli)≤80%CKSNAP; PCPseDNC; PSTNPss; DNA strandEnsemble learning5-fold cross-validation; Independent test0.984
    37.iPSW(2L)-PseKNC[45]3382 (E.coli)≤85%NCP; ANFSVM5-fold cross-validation0.905
    38.deepPromoter[46]3382 (E.coli)≤85%Combination of Continuous
    FastText N-Grams/ MRMD
    CNN5-fold cross-validation0.885
    39.iPSW(PseDNC-DL)[47]3382 (E.coli)≤85%one-hot; PseDNCCNN5-fold cross-validation0.925
      PWM: position weight matrix; SIDD: stress-induced DNA duplex destabilization; PCSF: position-correlation scoring function; ID: increment of diversity; PSTNPss: position-specific trinucleotide propensity based on single-strand; PseEIIP: electron-ion interaction pseudo-potentials of trinucleotide; PseKNC: pseudo k-tuple nucleotide composition; ANF: accumulated nucleotide frequency; PseZNC: pseudo multi-window Z-curve nucleotide composition; KNN: k-nearest neighbors; DAC: dinucleotide-based auto-covariance; PCPseDNC: parallel correlation pseudo dinucleotide composition; NCP: nucleotide chemical property; PseDNC: pseudo dinucleotide composition; mRMR: minimum redundancy maximum relevance; IFS: incremental feature selection; MRMD: maximum-relevance-maximum-distance; ANN: artificial neural network; SVM: support vector machine; FLD: fisher linear discriminant; SLP: single-layer perceptron; LSSVM: least square support vector machine; MD: mahalanobis discriminant; PLS: partial least squares; HMM: hidden markov models; RF: random forest; LR: logistic regression; CNN: convolution neural network; LDA: linear discriminant analysis.
  • 原核生物RNA聚合酶中的σ因子可以特异性识别并结合启动子,如图1所示。

    2005年至今已经开发了30多种计算方法来预测原核生物启动子,大致流程如图2所示。这些方法在许多方面有所不同,包括使用的基准数据集、特征提取方法、特征选择技术以及分类方法等。本文总结了39种原核启动子预测方法,从基准数据集信息、特征表示、特征选择、性能评估策略等多方面进行了比较和分析,如表1所示。

    39个预测工具根据其功能可分为以下3类。

    1)普通启动子的识别。工具1~9[9-17]属于这一类,这些工具收集各种原核生物的启动子作为基准数据集,包含大肠杆菌、枯草芽孢杆菌、结核杆菌、乳酸乳球菌、天蓝色链霉菌、分枝杆菌以及假单胞菌等。并没有指出这一类启动子具体的类型,因此这些方法只是简单地对启动子序列进行预测。

    2)特殊类型启动子的预测。这一类方法包含工具10~30[18-38]。这些工具以具体类型的启动子作为基准数据集,如大肠杆菌的6类启动子,原核生物的σ54启动子,蓝细菌的5类启动子等。不同类型的启动子在基因表达调控过程中起着不同且重要的作用,如目前已知的σ54启动子仅有数百条,而原核生物有3万多种,还有大量σ54启动子未被发现。σ54启动子参与了氮代谢的调控,因此σ54启动子的预测对于了解原核生物氮代谢过程具有重要意义。

    3)启动子的预测与分类。剩余的9个方法[39-47]均属于这一类,以大肠杆菌启动子作为数据集。这类方法具有一个典型的特征,即模型具有两层结构,第一层均是对启动子的预测,第二层是对启动子属性分类。工具31~36除了预测启动子和非启动子,第二层还判断启动子的具体类型(σ70, σ54, σ38, σ32, σ28, σ24)。实际上,启动子还有强弱之分。强启动子能增加转录频率从而提高基因的表达水平,所以预测启动子的强度也很重要。基于此,模型37~39的第二层鉴定启动子的强弱(Strong, Weak)。

    随着后基因组时代的到来以及计算机的发展,对于原核启动子的预测方法也不局限于初步的分类,还增加了对启动子类型和强度的鉴定,为了解基因调控过程提供新信息。

  • 建立原核启动子预测模型的第一步需要构建一个高质量的基准数据集。大肠杆菌(E.coli)作为原核生物中被广泛使用、研究的模式生物,其经过实验验证的转录调控信息已被系统地收录在RegulonDB数据库[48]中。DBTBS数据库[49]则收集整理了关于枯草芽孢杆菌(B.subtilis)的启动子数据。因此,RegulonDB和DBTBS数据库为预测方法提供了数据基础。39个工具中共有35个工具的数据集包含大肠杆菌和枯草芽孢杆菌启动子。

    另外,为了减少由序列同源性引起的潜在误差,通常会使用CD-HIT[50]工具以75%~85%的序列相似性阈值来去除掉数据集中序列冗余。原核启动子相较真核启动子,其结构相对较为简单、功能元件也相对较少,因此一般选择转录起始位点(transcriptional start site, TSS)上游60 bp以及下游20 bp作为原核启动子序列,不仅包含了重要的共有序列,如−35区、−10区、起始位点等,也避免了序列过长导致引入不必要的信息,具体数据可见原核启动子数据库(prokaryotic promoter database, PPD)[51]

  • 几乎所有的机器学习方法是以数值向量作为输入,因此需要一个合适的特征描述方法将数据集中的每一个样本转换为能够反映序列信息的数值向量。在原核启动子识别工作中,这些特征大致可以分为5类:核苷酸组成、核苷酸理化性质、伪核苷酸组成、二进制编码以及位置权重矩阵,以下对这5类特征进行简单的介绍。

  • 核苷酸组成,也叫k-mer,统计了DNA序列片段的所有可能组合的k长度子串出现频率,其计算公式为:

    式中,i代表某一k联体,有4k种可能性;N(t)表示DNA序列中某一k联体出现的次数;L表示DNA序列的长度。随着k值的增加,DNA序列的局部或短程信息也会逐渐增加。

    此外,核苷酸组成还包括了g-gapped k-mer,GC含量,累积核苷酸频率(accumulated nucleotide frequency, ANF)等。ANF表示了每一个碱基在序列中的分布密度,表达式为:

    式中,$ \left|{s}_{i}\right| $代表第i个碱基的位置;$ N\left({s}_{i}\right) $表示某一碱基出现频数;$ q\in \left\{A,C, G, T\right\}$

  • DNA序列中碱基的理化性质也可作为启动子预测的重要特征,包括核苷酸的化学性质、双链的稳定性、自由能、应激诱导的DNA双链不稳定性等。

    根据表2中对不同核苷酸的分类,DNA序列中第i个核苷酸可以表示为:

    Chemical propertyClassNucleotides
    Ring StructurePurineA, G
    PyrimidineC, T
    Functional GroupAminoA, C
    KetoG, T
    Hydrogen BondStrongC, G
    WeakA, T

    式中,xi, yi, zi分别表示指环结构(ring structure),功能组别(function group),以及氢键(hydrogen bond),如:

    因此4种碱基(A, C, G, T)可以分别表示为(1,1,1),(0,1,0),(1,0,0)和(0,0,1)。

  • 伪核苷酸组成(pseudo k-tuple nucleotide composition, PseKNC)最初是由文献[52]提出,分为I型和II型。这两种方法基于核苷酸的物化性质引入了DNA序列的全局或长程顺序信息。

    I型PseKNC,也叫平行相关伪核苷酸组成,将每一条DNA序列转化为4k + λ维的向量,具体表示为:

    II型PseKNC,也叫串联相关伪核苷酸组成,可产生4k + λ$ \Lambda $维向量:

    式(5)和式(6)中的$ {f}_{u} $与式(1)相同;前4k个元素是核苷酸组成特征,后面的元素是伪核苷酸组成特征;$ \mathrm{\lambda } $是一个正整数,反映序列顺序关联阶数;$ \omega $是权重因子,用于权衡核苷酸组分和DNA序列局部结构性质的影响;$ {\tau }_{j} $代表的是m阶关联因子,反映了每条DNA序列所有二核苷酸的m阶顺序关联性。

  • 二进制编码通过将4种核苷酸转换成包含4个元素的向量作为特征,其中一个元素为1,其余为0,既A、C、G和T分别表示为(1,0,0,0),(0,1,0,0),(0,0,1,0)以及(0,0,0,1)。因此,一段长为L的DNA序列可以用L×4的二维矩阵表示。

  • 位置权重矩阵(position weight matrix, PWM)可用来表示序列的保守片段,以序列每一位置的碱基保守程度为参量,分别计算每种碱基的保守指数,以此作为特征,具体表示为:

    式中,$ {S}_{i,j} $表示碱基i在第j个位置的保守指数;$ {q}_{_{i,j}} $是指在背景序列中碱基i出现在第j个位置的频率;$ {b}_{i} $是背景概率。

    因此,PWM是一个4×L的二维矩阵:

  • 从式(1)以及式(5)、式(6)可以看出,随着k值的增加,特征维度呈指数级增长,会导致“维度灾难”以及过拟合问题,而且由不同特征提取方法整合形成的融合特征集合往往会夹杂一些冗余或不相关的信息,所以为了避免出现上述问题并且提高计算效率,筛选有用的特征也是必不可少的步骤。

  • 最小冗余最大相关(minimum redundancy maximum relevance, mRMR)[53]是一种通过筛选相关性最大的特征来减少信息冗余的方法。mRMR的应用大大减少了特征维数和模型训练的时间,几乎不丢失有效信息。

    对于两个随机变量xy,其互信息为:

    式中,p()表示概率密度函数。

    最大相关性为:

    式中,c为类别变量;S为特征子集。

    最小冗余度则表示为:

    最后的评选标准如式(12)所示:

    mRMR会将所有特征的最大相关最小冗余打分按从大从小排序,值越大表明该特征越重要。

  • 当两个特征高度依赖时,它们对模型的贡献不能叠加,文献[54]基于距离函数提出了最大相关最大距离(max-relevance-max-distance, MRMD)来衡量每个特征的独立性。

    MRMD包含两个方面的特征排序度量:1)特征子集与目标类别的相关性;2)特征子集的冗余度。采用皮尔逊相关系数来衡量相关性、多种距离函数来计算冗余度。皮尔逊相关系数越大,特征与目标类别之间的相关性越高;特征距离越大,特征子集的冗余度越低;相关性与距离之和大的特征被选入最终的特征子集。因此,MRMD生成的特征子集冗余度最低,与目标类别的相关性最强。

  • F-score是一种基于filter的特征选择方法,对每一个特征进行重要性打分,其具体计算方法为:

    式中,$ {n}^{+}{\text{、}}{n}^{-} $分别表示正负样本的数量;$ {\bar{x}}_{i}^{\left(+\right)}{\text{、}} $$ {\bar{x}}_{i}^{\left(-\right)}{\text{、}}{\bar{x}}_{i} $分别指第i个特征在正样本、负样本以及所有样本中的平均值;$ {x}_{k,i}^{\left(+\right)}{\text{、}}{x}_{k,i}^{\left(-\right)} $分别指的是正负样本中第k条序列的第i个特征的数值。

    F-score通常与增量特征选择技术相结合来确定最优特征子集。

  • 增量特征选择(incremental feature selection, IFS)方法适用于确定最优特征子集。该方法的核心思想是将按重要性评分降序的特征依次加入到特征子集中,形成新的子集,将每一个子集输入至模型中,从而根据结果决策出最优特征子集。

  • 选择合适的算法可以使最终的模型具有良好的性能和泛化能力,各种监督学习方法已经被广泛应用于预测原核启动子,大致有以下4类。

  • 支持向量机(support vector machine, SVM)[55]是基于监督学习方式对数据进行二元分类,在样本空间中寻找最优分类超平面使得两类的间隔最大。

    对于线性可分的情况,存在一个分类超平面能将训练样本正确分类。而对于线性不可分的情况,需要使用核函数将低维不可分样本映射到更高维的特征空间,使得样本在高维空间中线性可分。

  • 神经网络(neural networks, NN)学习是一种模拟生物大脑神经网络的自适应计算模型。随着近年来人工智能的快速发展,人工神经网络(artificial neural network, ANN)及其卷积神经网络(convolutional neural network, CNN)已成为研究生物信息学问题的重要方法。

    基本的ANN结构包括输入层、隐藏层和输出层,主要特点是信号正向传播,误差反向传播。通过最小化误差函数,修正神经元间的连接权重,当其误差小于一定阈值的时候,即停止训练。

    CNN目前在很多研究领域都取得了巨大的成功,如语音识别、图像识别、自然语言处理等,是深度学习的代表算法之一。CNN通常由输入层、卷积层、激活函数、池化层、全连接层和输出层组成。与传统的神经网络不同的是CNN采用局部连接和权值共享,使得网络易于优化并且降低了模型的复杂度,减小过拟合风险。

  • 集成学习(ensemble learning, EL)通过构建并结合多个学习器来完成学习任务。在预测原核启动子的方法中,集成学习也是被广泛应用的,如随机森林(random forest, RF)。

    RF是一种基于决策树的集成学习方法,在决策树的训练过程中引入了随机属性选择。对于基决策树的每个结点,随机选择该结点属性集合中的一个子集,再从这个子集中选择一个最优属性用于划分。RF的每一个决策树都会产生一个分类结果,通过投票决定最终输出。与单一的决策树相比,RF具有较强的鲁棒性,并且对大数据具有较好的处理效果。

  • 线性判别分析(linear discriminant analysis, LDA)在二分类问题上最初是由文献[56]提出的,亦称为“Fisher判别分析”。

    LDA的核心思想相对简单:首先将训练集中的样本投影到一条直线上,使得同一类样本尽可能靠近,不同类样本尽可能远离;当新样本进来时,将其投影到同一直线上,从而根据投影点的位置判断其类别。

  • 在统计分析中,独立测试集和K折叠交叉验证已经被广泛地应用于验证分类器性能。当样本数量足够多时,会将基准数据集划分为训练集和独立测试集。独立测试集由于未参与模型的训练,可以更好地评价模型性能。在原核启动子识别模型中,K折叠交叉验证的应用最为广泛,其基本思想是重复利用数据,每一个样本既可以作为训练集参与模型训练,也会作为测试集参与模型评估。方法是将数据平均分成K份,K−1个子集用作训练,剩余一份用作测试,重复K次,最后返回K次结果的平均值。K折叠交叉验证最大程度上利用了每一个数据,能更好地反应模型的预测性能。

    另外,受试者工作特征曲线(receiver operating characteristic curve, ROC)下面积AUC值也可以反应模型性能,其值越接近于1,表明模型性能越好。

  • 近年来,基于生物信息学的原核启动子预测方法备受学者关注,已有多种方法被提出。为了充分了解这个领域的发展现状,本文收集并系统地分析了2005年至今共计39个原核启动子预测方法,详细阐述了这些方法的数据集构建、特征选择、特征提取、分类算法以及性能评估,详细信息如表1所示。

    目前,对原核启动子预测的研究取得了令人满意的结果。随着更多原核生物的基因组被测序出来,被研究的物种也不局限于少数几个模式生物,使用这些预测算法有助于了解原核生物基因调控机制。本文系统地比较了原核启动子预测方法,为研究此问题提供新思路、新角度。

Reference (56)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return