Thumbnails:
List:
Year:
Category:
Session:
Poster:
Getting poster data...
Daniel Gusenleitner, John Quackenbush, Aedin Culhane (Dana Farber Cancer Institute, Boston, USA)
Ten years ago, a seminal article by Hanahan and Weinberg described six hallmarks of cancer, intrinsic properties that drive the development and progression of disease. Given the wealth of gene expression profiling data, we investigated if we could computationally discover these properties using a purely data driven approach and determine if these recapitulate existing models of cancer. We performed a large scale meta-analysis of 14,103 gene expression profiles, which represented 22 tissue types. To derive associations between biological processes and phenotypes, GSEA was used on gene sets from Gene Signature Database and Molecular Signature Database. These associations were clustered using iBiBi, a novel bi-clustering approach, to extract modules of clinical covariates strongly associated with ranked sets of gene signatures. The analysis identified 13 multi-cancer modules corresponding to both new and well-known biological signals, including elevated proliferation in high grade tumors and an increased inflammatory micro-environment in cancerous tissues. The abundance of information contained in these modules, was represented using both established and new methods of visualization.