Date on Master's Thesis/Doctoral Dissertation

5-2022

Document Type

Doctoral Dissertation

Degree Name

Ph. D.

Department

Pharmacology and Toxicology

Degree Program

Pharmacology and Toxicology, PhD

Committee Chair

Frieboes, Hermann

Committee Co-Chair (if applicable)

Stienbach-Rankins, Jill

Committee Member

Stienbach-Rankins, Jill

Committee Member

El-Baz, Ayman

Committee Member

Rouchka, Eric

Committee Member

Hood, Joshua

Author's Keywords

cancer; mathematical modeling; metabolomics; chemotherapy; machine-learning

Abstract

Cancer is a complex and broad disease that is challenging to treat, partially due to the vast molecular heterogeneity among patients even within the same subtype. Currently, no reliable method exists to determine which potential first-line therapy would be most effective for a specific patient, as randomized clinical trials have concluded that no single regimen may be significantly more effective than others. One ongoing challenge in the field of oncology is the search for personalization of cancer treatment based on patient data. With an interdisciplinary approach, we show that tumor-tissue derived metabolomics data is capable of predicting clinical response to systemic therapy classified as disease control vs. progressive disease and pathological stage classified as stage I/II/III vs. stage IV via data analysis with machine-learning techniques (AUROC = 0.970; AUROC=0.902). Patient survival was also analyzed via statistical methods and machine-learning, both of which show that tumor-tissue derived metabolomics data is capable of risk stratifying patients in terms of long vs. short survival (OS AUROC = 0.940TEST; PFS AUROC = 0.875TEST). A set of key metabolites as potential biomarkers and associated metabolic pathways were also found for each outcome, which may lead to insight into biological mechanisms. Additionally, we developed a methodology to calibrate tumor growth related parameters in a well-established mathematical model of cancer to help predict the potential nuances of chemotherapeutic response. The proposed methodology shows results consistent with clinical observations in predicting individual patient response to systemic therapy and helps lay the foundation for further investigation into the calibration of mathematical models of cancer with patient-tissue derived molecular data. Chapters 6 and 8 were published in the Annals of Biomedical Engineering. Chapters 2, 3, and 7 were published in Metabolomics, Lung Cancer, and Pharmaceutical Research, respectively. Chapters 4 has been accepted for publication at the journal Metabolomics (in press) and Chapter 5 is in review at the journal Metabolomics. Chapter 9 is currently undergoing preparation for submission.

Share

COinS