Random matrix theory and covariance matrix filtering for cancer gene expression

Leif E. Peterson, Charles E. Ford

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

4 Scopus citations

Abstract

We investigated random matrix theory (RMT) and covariance matrix filtering with shrinkage techniques to characterize eigendecomposition of a 190 ×190 covariance matrix based on 750 genes and 18 tumor classes. Principal component subtraction using the first PC resulted in the most favorable outcome concerning eigenvector participation ratios, class-specific influence scores, and unsupervised clustering of arrays. By fitting the Marčenko-Pastur density function, we determined that 86.8% of the covariance matrix eigenvalues were below the threshold value of λ + = 0.5025, suggesting that they reside in the noise region. Removal of noise eigenvector effects in the data were not as informative as removal of only the first eigenvector, however, there were interesting properties observed among the 25 non-zero eigenvalues after noise removal - mostly that they were lower than the first 25 eigenvalues of the remaining types of covariance matrices.

Original languageEnglish (US)
Title of host publicationComputational Intelligence Methods for Bioinformatics and Biostatistics - 9th International Meeting, CIBB 2012, Revised Selected Papers
Pages173-184
Number of pages12
DOIs
StatePublished - 2013
Event9th International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, CIBB 2012 - Houston, TX, United States
Duration: Jul 12 2012Jul 14 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume7845 LNBI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other9th International Meeting on Computational Intelligence Methods for Bioinformatics and Biostatistics, CIBB 2012
Country/TerritoryUnited States
CityHouston, TX
Period7/12/127/14/12

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Random matrix theory and covariance matrix filtering for cancer gene expression'. Together they form a unique fingerprint.

Cite this