Squamous cell cancer of the oral cavity is associated with high case-fatality and high degree of disease and/or treatment-related orofacial functional impairment and disfigurement. Metastasis to the lymph nodes in the neck is the single most important independent predictor of survival from this disease and occurs in ~40% of patients. The diagnosis of metastasis relies on physical examination of the neck and auxiliary imaging using computed tomography, with or without positron emission tomography or magnetic resonance imaging. The current standard of care for patients with a tumor =2 cm without clinically apparent neck metastasis is close observation without prophylactic neck dissection to evaluate the presence of metastasis in cervical lymph nodes. For patients with larger tumors (>2 cm at the longest dimension), a prophylactic neck dissection is recommended. Unfortunately, a portion of patients without neck dissection develop lymph node metastases, and 15-60% who have a neck dissection do not have lymph node involvement. Thus, under the current clinical practice guidelines, substantial proportions of oral cancer patients are either under treated or over treated, pointing to the need to develop a means to accurately stratify patients according to their likelihood of having metastases. Using advanced metabolomics profiling methods and a well-characterized biorepository, this project will test the hypothesis that a saliva-based metabolite profile can distinguish patients with squamous cell oral cavity cancer who do and do not have lymph node metastases. In Aim 1, we will conduct global metabolomics and lipidomic profiling using LC-QTOF-MS, GC-MS and NMR on saliva samples from 90 oral cavity cancer patients with and without lymph node metastasis and 20 controls without cancer, all of them were enrolled in a prospective study of oral cancer in Seattle, Washington (NIH RO1 CA 09541, PI: Chen). (The control samples will be used to exclude metabolites showing large variations based on age, sex, race, and BMI.) In Aim 2, we will choose ~20-50 metabolites with concentrations that are found to differ in Aim 1 to a large degree between node-positive and node-negative patients to develop a reproducible and quantitative targeted assay using LC- MSMS and the same 90 samples; develop statistical models for the prediction of nodal metastasis in these same 90 patients; and validate the models using saliva samples from an independent set of 87 oral cavity cancer patients from the same parent study. This project will lay the ground work for the development of non- invasive tools to classify patients based on their lymph node metastasis status. Such tools have the potential to increase the chance of detecting metastasis early while reducing unnecessary surgical exploration among some oral cancer patients. It will set the stage for future large-scale RO1-type studies to validate our findings across population groups and to identify key modifying factors that may increase the accuracy and applicability of the predictive models.