Superresolution MUSIC based on Marčenko-Pastur limit distribution reduces uncertainty and improves DNA gene expression-based microarray classification

Leif E. Peterson

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We introduce a bootstrap root MUSIC (BRM) technique, which employs superresolution multisignal classification to reduce high-dimensional sets of genes from expression microarrays to low-dimensional sets used in supervised classification analysis. During BRM, the Marčenko-Pastur limit distribution of eigenvalues for the array-by-array gene expression covariance matrix was used for determining the eigenvalue cutoff for the noise subspace. Classifier results were compared with and without replacing gene expression values with the inverse of the distance to class-specific noise eigenspace for each microarray. Nine gene expression datasets were used for classification, and results of using BRM were compared with classification results based on use of random and best ranked N genes. On average, BRM resulted in greater classification of randomly selected genes when compared with direct use of randomly selected genes for classifier input. In addition, when BRM was applied to best ranked N genes, the interquartile ranges of accuracy were smaller when compared with direct input of best ranked genes into classifiers. Overall, BRM can optimally be used with 128 or 256 best ranked markers, requiring less extensive filtering to identify smaller sets of predictors. Use of a larger set of markers with BRM can help minimize the effect of concept drift over time.

Original languageEnglish (US)
Title of host publicationComputational Intelligence Methods for Bioinformatics and Biostatistics - 10th International Meeting, CIBB 2013, Revised Selected Papers
PublisherSpringer-Verlag
Pages194-209
Number of pages16
ISBN (Print)9783319090412
DOIs
StatePublished - Jan 1 2014
Event10th International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, CIBB 2013 - Nice, France
Duration: Jun 20 2013Jun 22 2013

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8452 LNBI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other10th International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, CIBB 2013
CountryFrance
CityNice
Period6/20/136/22/13

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Superresolution MUSIC based on Marčenko-Pastur limit distribution reduces uncertainty and improves DNA gene expression-based microarray classification'. Together they form a unique fingerprint.

Cite this