Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.

Non-WSU users: Please talk to your librarian about requesting this dissertation through interlibrary loan.

Access Type

WSU Access

Date of Award

January 2018

Degree Type

Dissertation

Degree Name

Ph.D.

Department

Computer Science

First Advisor

Sorin Draghici

Abstract

Currently, most diseases are diagnosed only after disease-associated changes have occurred. In this PhD dissertation, we propose a paradigm shift from treating the disease to maintaining the healthy state. The proposed approach is able to identify when systemic qualitative changes in biological systems happen, thus opening the possibility of therapeutic interventions before the occurrence of symptoms. The change detection method exploits knowledge from biological networks and longitudinal data using a system impact analysis approach. This approach is validated on eight datasets, for seven different model organisms and eight biological phenomena. On these data, our proposed method performs well, consistently identifying the qualitative change in each dataset. Most importantly, the method accurately detected the transition from the control stage (benign) to the early stage of hepatocellular carcinoma on an eight-stage disease dataset. Knowing when a transition (qualitative change) from healthy to disease occurs may help preserve the healthy state.

We also propose a novel analysis approach for metabolic pathway analysis that uses an impact analysis approach and leverages the stoichiometry of bio-chemical reactions to identify which pathways are significantly disrupted by the change in metabolite levels in disease samples versus healthy controls. Our approach outperforms the over-representation approach when evaluated on simulated data. We applied our proposed method to biological experiment data that compares samples from pregnant women to non-pregnant control samples. Our method was able to identify biologically relevant results on real high-throughput data better than the classical approach.

In summary, we developed two novel methods for the analysis of high-throughput biological data, gene expression and metabolite concentration, respectively. The proposed methods can be adapted to work together in order to capture relevant complementary information stored in time-course datasets for gene expression or metabolite levels that may available for complex diseases in order to identify when a qualitative change happens, before the physiological onset of the disease.

Off-campus Download

Share

COinS