Detection of Statistically Significant Sub-Clusters in Biological Data

• Activity Number: 194
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Guoli Sun*+ and Alexander Krasnitz
• Companies: Stony Brook University and Cold Spring Harbor Laboratory
• Keywords: hierarchical clustering ; significant sub-clusters ; biological data ; permutation tests ; pvalue estimation ; detection of subtypes

Abstract

TEST (Trees Evaluated Statistically for Tightness) is a computational approach to finding tight branches in hierarchical trees, complete with statistical assessment of findings. Multiple data sets of different biological origins, including mRNA expression, protein expression and DNA copy number variation, are used to validate our approach. When applied to these benchmark cases, our procedure outperforms published methods. As an application, we use a massive DNA copy number variation dataset for ovarian serous carcinoma to derive four sub-classes of the disease. These exhibit strikingly different profiles of mRNA expression, patterns of DNA methylation and clinical prognoses.