Transcriptomic network analysis of brain aging and alzheimers disease


Thesis Type: Postgraduate

Institution Of The Thesis: Middle East Technical University, Turkey

Approval Date: 2017

Thesis Language: English

Student: Poorya Parvizi

Co-Consultant: NURCAN TUNÇBAĞ, MEHMET SOMEL

Abstract:

Multiple studies have investigated aging brain transcriptomes to identify for age-dependent expression changes and determine genes that may participate in age-related dysfunction. However, aging is a highly complex and heterogeneous process where multiple genes contribute at different levels depending on individuals’ environments and genotypes. Both this biological heterogeneity of aging, as well as technical biases and weaknesses inherent to transcriptome measurements, limit the information gained from a single data set. Here we propose using network analysis to reproducibly identify aging-related gene interactions shared across different datasets. We employ the prize-collecting Steiner forest algorithm to create aging networks on human brain transcriptome datasets. The algorithm calculates the optimal interaction set among aging-related genes within a protein-protein interaction (PPI) network, taking into consideration expression-age correlation coefficients of the most deferentially expressed genes with age, and the PPI confidence scores. This allows aging-related genes to interact either directly or through intermediate nodes. The intermediate nodes, in turn, can represent genes undetected in transcriptome data due to low light intensity, technical inefficiency of platforms, or aging-related molecular changes that do not involve mRNA abundance change, such as aging-related post-translational modifications. Using the predicted networks, we have performed network alignment of the reconstructed networks to test whether common interactions might be found in different tissues’ aging networks. In addition, we also extend the approach to compare molecular changes during aging and in Alzheimer’s Disease. We hypothesize that using network alignment will help highlight the most relevant gene clusters and pathways shared between the two processes.