A framework for gene co-expression network analysis of lung cancer


Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Enformatik Enstitüsü, Sağlık Bilişimi Anabilim Dalı, Türkiye

Tezin Onay Tarihi: 2013

Öğrenci: ERHAN AKDEMİR

Eş Danışman: YEŞİM AYDIN SON, TOLGA CAN

Özet:

Construction method of a gene co-expression network (GCN) is crucial in medical research aiming to reveal disease related genes. Applied similarity measure and selection of edges that represent significantly co-expressed gene pairs in the network affect directly the elements of a network and so the list of prioritized genes. Pearson correlation coefficient is a commonly used similarity measure to quantify co-expressions of genes due to its simplicity and performance compared to many complex methods. However, it is affected by outliers and may not be reliable with low sample size. On the other hand, selection of edges is generally based on an arbitrary cutoff which makes networks subjective. For a more standard and accurate analysis, reliability of a similarity measure must be ensured as well as an objective threshold determination for the selection of edges. Here, a framework is proposed for the construction of GCNs that combines a reliability measure, stability, previously applied to Pearson correlation coefficient to detect general co-expression differences between healthy and cancer state and an automatic threshold selection method, Random Matrix Theory for a standard network construction. The proposed framework was applied to lung adenocarcinoma. In the analysis part, genes were prioritized by using changes in topological and neighborhood properties of nodes in control and disease networks. Differential co-expressions of known interacting proteins and intrinsically disordered proteins were also analyzed. Results suggest that co-expression networks are topologically spoke-like and control samples are in transition phase from healthy to cancer. Thus, effects of stability on finding general co-expression differences between cancer and healthy states could not be assessed. Prioritized genes by both proposed and control methods are mostly enriched to relevant processes reflect the changes in cellular machinery as a result of a state shift to cancer and may reveal dynamical features of transition of cells to cancer state with a further detailed analysis. Furthermore, some genes were prioritized related with cilia which may have roles early phases of transition.