Despite consistently appearing as a top 5 cause of death domestically and internationally, understandings of the underlying biological interactions of COPD remain insufficient. Particularly, early identification of highly vulnerable patients that may experience more severe infectious exacerbation is a challenge during diagnostics. This project leverages unsupervised clustering and multi-resolution decomposition of proteomic abundance profiles in a small cohort of COPD subjects to identify candidate biomarkers of heightened exacerbation risk. As part of an ongoing study, serum and sputum samples were collected during well visits (stable illness) from n=67 subjects, and analyzed using quantitative proteomics. A scores principal component analysis was performed with simple value decomposition to visualize the feature space of the protein dataset at different resolutions, with the goal of correlating a specific cluster and subset of proteins directly to patient severity. Subsequently, the loading values utilized by the analysis were retrieved and used to identify a set of 15 serum proteins with the highest association to clinical outcome and severity. Finally, an additional bimodal clustering technique was used to separate frequent and infrequent exacerbation thresholds within our patient cohort. Through validation of these top 15 proteins with ontology and network analysis, we reveal this protein set as a likely clinically significant group of biomarkers which may bolster both diagnostics and treatment approaches in the field.

Publication Date


Document Type


Student Type


Degree Name

Bioinformatics (MS)

Department, Program, or Center

Thomas H. Gosnell School of Life Sciences


College of Science


Stefan Schulze

Advisor/Committee Member

Gordon Broderick

Advisor/Committee Member

Gary Skuse


RIT – Main Campus