Over the past decade, precision medicine research in head and neck cancer (HNC) has been singularly focused on a single biomarker, human papillomavirus (HPV), to target low-risk patients for treatment de-intensification (i.e., reduce therapy to improve quality of life without sacrificing survival). De-intensification is vital since standard treatment involves toxic doses of chemoradiation that causes adverse events and debilitating long-term side effects in most patients. In 2019, however, two phase-III clinical trials confirmed that HPV was insufficient to target the most promising de-intensified therapy. This failure presents a timely opportunity to test new prognostic markers to target de-intensified therapy in the 41 existing clinical trials, but a lack of promising biomarkers hinders progress. HNC is a disease marked by extensive metabolic heterogeneity that is emerging as a rich source of novel prognostic biomarkers but has yet to be fully investigated. For instance, many of the somatic mutations in HNC that are associated with prognosis (e.g., TP53 and PIK3CA) are upstream regulators of key metabolic processes in glycolysis, lipids, amino acids, and cell signaling. Recent studies have showed that the Warburg effect, the atypical metabolic state in which a tumor upregulates glycolysis but bypasses mitochondrial oxidative phosphorylation (OXPHOS), is a key discriminating characteristic between HNC cell lines linked to prognosis. Most recently, five different studies in 2020 found that gene expression probes targeting metabolic regulating genes created signatures that can predict HNC patient survival. The evidence leads to one conclusion— metabolism is intimately involved in the development, progression, and survival of HNC. Thus, metabolomics may be the best approach to discover the translational biomarkers vital to the future of HNC precision medicine. However, the data required to investigate and validate an HNC metabolomic biomarker, particularly in a clinical setting, is severely lacking. My program of research aims to fill that gap. In my KL2 research, which includes 209 HNC patients with blood plasma metabolomics, I used machine learning algorithms to identify a pretreatment metabolomic endotype comprised of glycolysis, OXPHOS, lipid biosynthesis, and amino acid metabolites. The endotypes categorized patients into one of two distinct groups that were associated with overall survival (hazard ratio = 2.39, 95% CI: 1.19-4.78). Most importantly, the prognostic ability of the endotype was independent of relevant clinical factors such as HPV, smoking, age, sex, race, tumor site and stage, suggesting that it may succeed as an HNC biomarker where others have failed. The goal of this R03 is to validate my KL2 findings with an independent cohort (Aim 1) and investigate how the metabolomic endotype in blood is related to metabolism and metabolic characteristics in the tumor (Aim 2). Positive findings from this study will help me translate this bi...