A Systems Biology Approach To Identify Dna Methylation Based Disease Biomarkers
Identifying a set of biomarkers that can accurately distinguish the patients from healthy individuals is one of the major challenges in medical science. Although several meta-analysis and multi-omics data integration approaches are currently available that aim to overcome this challenge, the majority of them either (i) do not integrate epigenetic data (i.e., DNA methylation) with other data types, or (ii) do not consider the complex ways in which genes interact. Methylation is known to play a vital role in the progression of many complex diseases such as cancer. In this thesis, we propose a novel computational framework that integrates (i) DNA methylation data, ii) gene expression data, and (iii) known pathway knowledge, in order to identify robust and reproducible diagnostic biomarkers. We illustrate our framework by identifying biomarkers for three important diseases: (i) colorectal adenocarcinoma, (ii) kidney renal clear cell carcinoma, and (iii) glioblastoma multiforme. We validate our panels of biomarkers by comparing them with the results of other three existing relevant computational approaches, as well as with 22 other established disease-specific signature panels using 1,994 additional samples obtained from 30 independent, validation-only datasets. The results show that the set of biomarkers identified by the proposed framework is able to distinguish the patients from healthy individuals with significantly higher sensitivity and specificity. compared to others.