Date of Award

4-4-2017

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Science

First Advisor

Coskun Bayrak

Abstract

Background The genome project had given us a brief view of what and how an organ is made up of. In 2000’s after the genome project, every year the FDA approved around 100 drugs and was an estimate that these numbers would grow up each year, unlike the numbers only decreased. Hence, it was concluded that these facts would not be necessary, as diseases were treated with respect to the organs effected. With a growing data of molecular interaction in a cell or tissue type, proteomic, genomic, and metabolic data, will the old classification still hold good? These large-scale data have been always seen as inter connected networks. Disease Similarity has been estimated based on clinical manifestation traditionally, whereas with availability of transcriptomic data the relationships between diseases must be revisited. While the current methods use mining of expression data, which is biased to well-studied diseases only. Methods A computational model is developed for measuring disease similarity using gene expression and molecular data. After obtaining data from different biological sources, we took benefit from the varied knowledge base of molecular networks to obtain novel associations between diseases. We analyzed the data at gene level and provided the similarity measure. For each disease at the pathway level, differential co-expression of each gene associated with the disease is calculated. Disease similarity was estimated as a coefficient between any two diseases. We also compare the results with existing Genome-Wide Association Study (GWAS) data and found that the associations correlate with them at a higher rate. Results We developed a network of human diseases with Disease-to-Disease (D2D) associations among various diseases using expression data and various measures on it. We compared the results with existing associations and found out that the disease links shared disease genes and drugs. It was also found that the similarity score is significantly higher confirming that our measures can be used to recover further associations.

Download

Share

COinS