Date of Award

12-4-2015

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Bioinformatics

First Advisor

Huixiao Hong

Abstract

The human leukocyte antigens (HLAs) can capture endogenous or exogenous peptides and set up a cascade which results in immune activation to fight a pathogen or sometimes to trigger autoimmune diseases or adverse drug reactions. Understanding HLA-peptide binding is essential for the development of vaccines and protein therapies, the identification of methods to prevent autoimmune diseases and the prediction of adverse drug reactions. Using an in silico approach we generated a 9-mer peptidome for Class I HLAs from the human proteome and harvested experimental HLA-peptide binding data from a variety of databases. Unfortunately, many peptides in the peptidome do not have experimental HLA binding data in the databases. To construct a complete HLA-binding peptidome, these will need to be predicted. After a literature review, we found the current prediction models for HLA-peptide binding have limitations and are not reliable for all the peptides in the absence of experimental data. Therefore, we developed a network-based algorithm to overcome the limitations and to provide HLA binding predictions for the peptides without experimental data. We found network analysis is an effective approach for analysis of sparse big datasets such as the HLA-binding of peptides. We developed sNebula, a similar neighbor-edges based and unbiased leverage algorithm and this outperformed the existing methods. We predicted Class I HLA-binding for the peptides without experimental data using sNebula and complete the construction of a Class I HLA-binding peptidome. We showed the Class I HLA-binding peptidome could significantly improve prediction of adverse drug reactions through binding to Class I HLA using molecular docking. Since the HLAs and peptides are two core components to trigger immunologic responses, the complete Class I HLA-binding peptidome accomplished in this study provides the scientific community a rich source for understanding the interactions between Class I HLAs and peptides, potentially assisting in the development of vaccines and protein therapeutics, facilitating studies of autoimmune diseases and improving the prediction of drug adverse reactions.

Download

Share

COinS