Melvin's digital garden

RielMehan2010

CREATED: 201001261315 LINK: url:~/Modules/Literature/RielMehan2010.pdf Title: An integrative modular approach to systemically predict gene-phenotype associations

** Challenges in microarray expression

  • normalization
  • multiple platforms
  • outliers

** Network based integration

  • co-expression network per microarray
  • find recurrent dense subgraphs in multiple co-expression graphs
  • want co-expression clusters that are enriched for specific functions/conditions
  • map dataset to phenotype term using unified medical language

** Algorithm

  • based on simulated annealing to find subset of vertices that are
  • large
  • dense
  • phenotype specific (hypergeometric p-value)

Use overrepresentation of GO terms to validate gene sets = power of gene sets for annotation

http://zhoulab.usc.edu

Links to this note