Assessing the Impact of Tumor Heterogeneity on Powering Microarray Signature Discovery

• Activity Number: 532
• Type: Contributed
• Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Timothy Scott Davison*+ and Sian Dibben and Janet Taylor and Robert J. Holt and Paul J. Kelly and Ian Paul and Peter Kerr and Dean A. Fennell and Jacqueline A. James and Richard D. Kennedy
• Companies: Almac Diagnostics and Almac Diagnostics and Almac Diagnostics and Almac Diagnostics and Belfast Health and Social Care Trust and Queen's University Belfast and Almac Diagnostics and Queen's University Belfast and Queen's University Belfast and Almac Diagnostics
• Keywords: Classification ; Tumor ; Signature ; Gene expression ; Heterogeneity

Abstract

Intratumoral variation at the gene expression level was evaluated to inform the design of a prognostic signature development study. 10 NSCLC patients were selected and 3-5 blocks were taken from each patient representing different parts of the tumor. RNA was extracted from both macrodissected and whole FFPE sections and processed on a lung cancer specific microarray. Expression data and simulated multi-gene signatures were fit to an ANOVA model to assess the statistical significance of observed variation. Results: Variation exists in gene expression between blocks taken from a single patient but is less than variation between patients and disease states. Macrodissection and sampling dependent multi-gene signatures reduce overall measurement error due to sample heterogeneity and the number of samples required to observe a fixed hazard ratio.