Analysis of gene expression profiles: class discovery and leaf ordering

  • Authors:
  • Chris H. Q. Ding

  • Affiliations:
  • University of California, Berkeley, CA

  • Venue:
  • Proceedings of the sixth annual international conference on Computational biology
  • Year:
  • 2002

Quantified Score

Hi-index 0.01

Visualization

Abstract

We approach the class discovery and leaf ordering problems using spectral graph partitioning methodologies. For class discovery or clustering, we present a min-max cut hierarchical clustering method and show it produces subtypes quite close to human expert labeling on the lymphoma dataset with 6 classes. On optimal leaf ordering for displaying the gene expression data, we present a sequential ordering method that can be computed in O(n2) time which also preserves the cluster structure. We also show that the well known statistic methods such as F-statistic test and the principal component analysis are very useful in gene expression analysis.