The incidence of esophageal adenocarcinoma (EAC) has risen 10-fold over the past half century and continues to have a dismal prognosis. Barrett’s esophagus (BE) is the precursor lesion to EAC, and patients diagnosed with BE undergo surveillance and endoscopic therapy for early neoplasia. However, more than 90% of EAC patients are never diagnosed with BE beforehand, and widespread upper endoscopy to identify patients with BE is expensive and of questionable value. Thus, there is an urgent need to develop minimally-invasive methods of BE screening that can be easily performed in the primary care setting to allow for efficient and cost-effective interventions to decrease EAC mortality. The esophageal microbiome is heavily influenced by migration of bacteria from the mouth via swallowed secretions. The esophageal microbiome is altered in gastroesophageal reflux and BE, and these changes may therefore reflect changes in the oral microbiome, an easily accessible sampling site. In fact, we have demonstrated that there are marked alterations to the oral microbiome in patients with BE, and a model based on specific taxa can distinguish BE patients with high sensitivity and specificity. We hypothesize that microbiome analysis of saliva can identify patients with BE with high accuracy, thus representing a novel, non-invasive screening test to identify patients at risk for EAC. In the current proposal we aim to validate our preliminary findings in a large endoscopic cohort. We propose to leverage a completed study of patients undergoing a first upper endoscopy with associated saliva samples, as well as patients with suspected early neoplasia (high grade dysplasia or early EAC). In Aim 1, we will determine whether the oral microbiome identifies patients with BE. We hypothesize that a microbiome score based on a model containing relative abundance of Lautropia, Streptococcus, and Enterobacteriaceae identifies patients with and without BE with high accuracy. In Subaim 1a, we will determine whether the oral microbiome in combination with a clinical prediction model (Michigan Barrett’s Esophagus pREdiction Tool; M-BERET) identifies patients with Barrett’s esophagus. As external validation of an oral microbiome signature, we will perform a case-control study using saliva collected as part of a nationwide, primary care-based BE screening study (Subaim 1b). In Aim 2, we will assess whether an oral microbiome signature can identify patients with BE and early neoplasia. In Aim 3, we will conduct a prospective cohort study of 250 patients with and without BE, and collect serial saliva samples to determine whether repeated sampling of the oral microbiome improves identification of patients with BE. We will also assess temporal stability of an oral microbiome signature for the diagnosis of BE. We propose a novel, biologically-based non-endoscopic approach to change the paradigm for EAC prevention. We hope that this will lead to development of a laboratory-based testing strategy that is highly acceptable to patients, is cost-effective, and can be easily translated to the primary care setting.