Gene Selection for Multiclass Prediction by Weighted Fisher Criterion

1687-4153-2007-64628 1687-4153 Research Article Gene Selection for Multiclass Prediction by Weighted Fisher Criterion XuanJianhuaxuan@vt.edu WangYueyuewang@vt.edu DongYibinyibin.dong@vt.edu FengYuanjianyjfeng@vt.edu WangBinbinwang@vt.edu KhanJavedkhanjav@mail.nih.gov BakayMariambakay@cnmcresearch.org WangZuyizwang@cnmcresearch.org PachmanLaurenpachman@northwestern.edu WinokurSarastwinoku@uci.edu ChenYi-Wenychen@cnmcresearch.org ClarkeRobertclarker@georgetown.edu HoffmanEricehoffman@cnmcresearch.org

Department of Electrical and Computer Engineering, Virginia Polytechnic Institute and State University, Arlington, VA 22203, USA

Department of Pediatric Oncology, National Cancer Institute, Gaithersburg, MD 20877, USA

Research Center for Genetic Medicine, Children's National Medical Center, Washington, DC 20010, USA

Disease Pathogenesis Program, Children's Memorial Research Center, Chicago, IL 60614, USA

Department of Biological Chemistry, University of California, Irvine, CA 92697, USA

Lombardi Cancer Center, Georgetown University, Washington, DC 20007, USA

EURASIP Journal on Bioinformatics and Systems Biology 1687-4153 2007 2007 1 64628 http://bsb.eurasipjournals.com/content/2007/1/64628 10.1155/2007/64628 30820061612200620320071072007 2007Jianhua Xuan et al.This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Gene expression profiling has been widely used to study molecular signatures of many diseases and to develop molecular diagnostics for disease prediction. Gene selection, as an important step for improved diagnostics, screens tens of thousands of genes and identifies a small subset that discriminates between disease types. A two-step gene selection method is proposed to identify informative gene subsets for accurate classification of multiclass phenotypes. In the first step, individually discriminatory genes (IDGs) are identified by using one-dimensional weighted Fisher criterion (wFC). In the second step, jointly discriminatory genes (JDGs) are selected by sequential search methods, based on their joint class separability measured by multidimensional weighted Fisher criterion (wFC). The performance of the selected gene subsets for multiclass prediction is evaluated by artificial neural networks (ANNs) and/or support vector machines (SVMs). By applying the proposed IDG/JDG approach to two microarray studies, that is, small round blue cell tumors (SRBCTs) and muscular dystrophies (MDs), we successfully identified a much smaller yet efficient set of JDGs for diagnosing SRBCTs and MDs with high prediction accuracies (96.9% for SRBCTs and 92.3% for MDs, resp.). These experimental results demonstrated that the two-step gene selection method is able to identify a subset of highly discriminative genes for improved multiclass prediction.

[1234567891011121314151617181920212223242526272829303132333435363738394041424344454647]

Molecular classification of cutaneous malignant melanoma by gene expression profilingBittnerMMeltzerPChenYNature2000406679553654010.1038/3502011510952317Molecular classification of cancer: class discovery and class prediction by gene expression monitoringGolubTRSlonimDKTamayoPScience1999286543953153710.1126/science.286.5439.53110521349Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learningShippMARossKNTamayoPNature Medicine200281687410.1038/nm0102-6811786909Molecular profiling of human cancerLiottaLPetricoinENature Reviews Genetics200011485610.1038/3504956711262874Statistical pattern recognition: a reviewJainAKDuinRPWMaoJIEEE Transactions on Pattern Analysis and Machine Intelligence200022143710.1109/34.824819Feature selection: evaluation, application, and small sample performanceJainAKZongkerDIEEE Transactions on Pattern Analysis and Machine Intelligence199719215315810.1109/34.574797Small sample size effects in statistical pattern recognition: recommendations for practitionersRaudysSJJainAKIEEE Transactions on Pattern Analysis and Machine Intelligence199113325226410.1109/34.75512FukunagaKIntroduction to Statistical Pattern RecognitionAcademic Press, Boston, Mass, USA21990DevijverPAKittlerJPattern Recognition: A Statistical ApproachPrentice-Hall, Englewood Cliffs, NJ, USA1982Floating search methods in feature selectionPudilPNovovicovaJKittlerJPattern Recognition Letters199415111119112510.1016/0167-8655(94)90127-9Comparison of discrimination methods for the classification of tumors using gene expression dataDudoitSFridlyandJSpeedTPJournal of the American Statistical Association200297457778710.1198/016214502753479248Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networksKhanJWeiJSRingnérMNature Medicine20017667367910.1038/89044128252111385503A comparative study of feature selection and multiclass classfication methods for tissue classification based on gene expressionLiTZhangCOgiharaMBioinformatics200420152429243710.1093/bioinformatics/bth26715087314Diagnosis of multiple cancer types by shrunken centroids of gene expressionTibshiraniRHastieTNarasimhanBChuGProceedings of the National Academy of Sciences of the United States of America200299106567657210.1073/pnas.08209929912444312011421Biomarker identification by feature wrappersXiongMFangXZhaoJGenome Research200111111878188731115011691853LoogMApproximate Pairwise Accuracy Criteria for Multiclass Linear Dimension Reduction: Generalisations of the Fisher CriterionDelft University Press, Delft, The Netherlands1999Multiclass linear dimension reduction by weighted pairwise Fisher criteriaLoogMDuinRPWHaeb-UmbachRIEEE Transactions on Pattern Analysis and Machine Intelligence200123776276610.1109/34.935849Generalized inverse of a singular matrixKoopJCNature196320071614109987PressWMFlanneryBPTeukolskySAVetterlingWTNumerical Recipes: The Art of Scientific ComputingCambridge University Press, New York, NY, USA1986A branch and bound algorithm for feature subset selectionNarendraPMFukunagaKIEEE Transactions on Computers1977269917922On the effectiveness of receptors in cognition systemMarillTGreenDMIEEE Transactions on Information Theory19639111710.1109/TIT.1963.1057810A direct method of nonparametric measurement selectionWhitneyAWIEEE Transactions on Computers197120911001103On selecting features for pattern classifiersStearnsSDProceedings of the 3rd International Conference on Pattern Recognition, Coronado, Calif, USA, November 19767175HaykinSNeural Networks: A Comprehensive FoundationPrentice-Hall, Upper Saddle River, NJ, USA21999Classification of multiple cancer types by multicategory support vector machines using gene expression dataLeeYLeeC-KBioinformatics20031991132113910.1093/bioinformatics/btg10212801874Multiclass cancer diagnosis using tumor gene expression signaturesRamaswamySTamayoPRifkinRProceedings of the National Academy of Sciences of the United States of America20019826151491515410.1073/pnas.2115663986499811742071Sources of variability and effect of experimental approach on expression profiling data interpretationBakayMChenY-WBorupRZhaoPNagarajuKHoffmanEBMC Bioinformatics200231410.1186/1471-2105-3-46569111936955Nuclear envelope dystrophies show a transcriptional fingerprint suggesting disruption of Rb-MyoD pathways in muscle regenerationBakayMWangZMelconGBrain20061294996101310.1093/brain/awl02316478798Statistical algorithms description documentAffymetrix Technical NoteAffymetrix2002http://www.affymetrix.com/support/technical/whitepapers/sadd_whitepaper.pdfIn vivo filtering of in vitro expression data reveals MyoD targetsZhaoPSeoJWangZWangYShneidermanBHoffmanEComptes Rendus - Biologies200332610-111049106510.1016/j.crvi.2003.09.03514744113Embryonic myogenesis pathways in muscle regenerationZhaoPHoffmanEDevelopmental Dynamics2004229238039210.1002/dvdy.1045714745964Expression profiling of FSHD muscle supports a defect in specific stages of myogenic differentiationWinokurSChenY-WMasnyPSHuman Molecular Genetics200312222895290710.1093/hmg/ddg32714519683A web-accessible complete transcriptome of normal human and DMD muscleBakayMZhaoPChenJHoffmanENeuromuscular Disorders2002121S125S14112206807Expression profiling in the muscular dystrophies: identification of novel aspects of molecular pathophysiologyChenY-WZhaoPBorupRHoffmanEJournal of Cell Biology200015161321133610.1083/jcb.151.6.1321219060011121445Dystrophin: the protein product of the Duchenne muscular dystrophy locusHoffmanEBrownRHJr.KunkelLMCell198751691992810.1016/0092-8674(87)90579-43319190Complete cloning of the Duchenne muscular dystrophy (DMD) cDNA and preliminary genomic organization of the DMD gene in normal and affected individualsKoeningMHoffmanEBertelsonCJMonacoAPFeenerCKunkelLMCell198750350951710.1016/0092-8674(87)90504-63607877Slug is a novel downstream target of MyoD. Temporal profiling in muscle regenerationZhaoPIezziSCarverEJournal of Biological Chemistry200227733300913010110.1074/jbc.M20266820012023284Microarray synthesis through multiple-use PCR primer designFernandesRJSkienaSSBioinformatics2002181S128S13510.1093/bioinformatics/18.suppl_1.S12812385994Early diagnostic marker panel determination for microarray based clinical studiesJaegerJWeichenhanDIvandicBSpangRStatistical Applications in Genetics and Molecular Biology200541, article 9How many genes are needed for early detection of breast cancer, based on gene expression patterns in peripheral blood cells?LiWBreast Cancer Research200575E510.1186/bcr1295124215316168099Converting a breast cancer microarray signature into a high-throughput diagnostic testGlasAMFlooreADelahayeLJMJBMC Genomics2006727810.1186/1471-2164-7-278163604917074082Classifiers in almost empty spacesDuinRPWProceedings of the 15th International Conference on Pattern Recognition (ICPR '00), Barcelona, Spain, September 2000217Evolution and generalization of a single neurone—I: single-layer perceptron as seven statistical classifiersRaudysSJNeural Networks199811228329610.1016/S0893-6080(97)00135-412662838Evolution and generalization of a single neurone—II: complexity of statistical classifiers and sample size considerationsRaudysSJNeural Networks199811229731310.1016/S0893-6080(97)00136-612662839Expected classification error of the Fisher linear classifier with pseudo-inverse covariance matrixRaudysSJDuinRPWPattern Recognition Letters1998195-638539210.1016/S0167-8655(98)00016-6VapnikVNStatistical Learning TheoryJohn Wiley & Sons, New York, NY, USA1998OjaESubspace Methods of Pattern RecognitionJohn Wiley & Sons, New York, NY, USA1984