Access Type

Open Access Dissertation

Date of Award

January 2020

Degree Type


Degree Name



Computer Science

First Advisor

Sorin Draghici


The traditional drug discovery process is extremely slow and costly. More than 90% of drugs fail to pass beyond the early stage of development and toxicity tests, and many of the drugs that go through early phases of the clinical trials fail because of adverse reactions, side effects, or lack of efficiency. In spite of unprecedented investments in research and development (R&D), the number of new FDA-approved drugs remains low, reflecting the limitations of the current R&D model.

In this context, finding new disease indications for existing drugs sidesteps these issues and can therefore increase the available therapeutic choices at a fraction of the cost of new drug development.

In this thesis, we introduce a drug repurposing approach that takes advantage of prior knowledge of drug targets, disease-related genes, and signaling pathways to construct a drug-disease network composed of the genes that are most likely perturbed by a drug. Systems biology can be used as an effective platform in drug discovery and development by leveraging the understanding of interactions between the different system components. By performing a system-level analysis on this network, our approach estimates the amount of perturbation caused by drugs and diseases and discovers drugs with the potential desired effects on the given disease. Next, we develop a stable clustering method that employs a bootstrap approach to identify the stable clusters of cells. We show that strong patterns in single cell data will remain despite small perturbations. The results, that are validated based on well-known metrics, show that using this approach yields improvement in correctly identifying the cell types, compared to other existing methods.