A new approach for biomarker detection using fusion networks

Abstract

In this thesis we propose a new approach for biomarker detection using single source and fusion networks. Our algorithm detects metastable regions with similar and high weights in networks. Standard methods for biomarker detection analyse the differentially expression, or other biological measurement, of genes, without taking any further biological knowledge into account. Network approaches include further insight into the analysis, nevertheless most of them are limited to the examination of one data type. With our fusion network algorithm we introduce a new and promising approach. We analyse breast cancer gene expression and methylation data using our fusion network approach. The proposed method detects known and novel biomarkers, which are highly supported by breast cancer literature, biological pathways, and classification power.