PROJECT SUMMARY Epilepsy is one of the world’s most prevalent diseases, yet the rate of uncontrolled seizures has not changed in decades. One of the reasons for this is our limited understanding of seizure mechanisms, and so one of the main goals of epilepsy research is to identify new biomarkers to help us understand the nature of the disease. Recent technological advancements now allow us to monitor brain activity with much higher resolution, which have led to the identification of promising potential biomarkers such as High Frequency Oscillations (HFOs). Unfortunately, clinicians still have not determined how to utilize this information under clinical conditions. There are three main obstacles to implementing HFOs in practice: 1) they are difficult to find; 2) it is unclear how to ascertain which HFOs are truly related to epilepsy; and 3) it is unclear how to use the HFO data in a prospective fashion to improve clinical care. The purpose of this project is overcome each of these obstacles. In the past funding period, we developed and validated an HFO detection algorithm that overcomes the first obstacle, and allowed us acquire a massive database of HFOs that have opened new avenues of research. In this proposal, we will leverage that algorithm to move HFOs towards clinical translation. In the first Aim, we apply advanced functional connectivity techniques to quantify the network properties of HFOs. Our data, which comprise HFOs from the entire hospitalization and fully curated metadata, are ideal for robust analyses of this new area of HFO research. The second Aim addresses a longstanding, and still unsolved problem in HFO research: how to discern when HFOs are due to epileptic processes versus normal physiology? Our past funding period identified some potential methods to identify pathological HFOs, but also crucial caveats that must be addressed prior to clinical implementation. This Aim will combine multiple classification methods with state-of-the-art machine learning tools to distinguish epileptic from normal HFOs. It will also conduct a large human expert classification of HFOs using clinical EEG software, to start involving epilepsy clinicians in the direct evaluation of HFOs. The third Aim will further develop the translational potential of HFOs, incorporating our unique longitudinal clinical data to characterize the effects of medications, sleep, and other time-varying effects on HFO rates. It will then incorporate these and all prior HFO data into a rigorous latent class model to predict how likely each channel is to be epileptic. These Aims together serve as the framework to establish HFOs as a clinically viable biomarker of seizures, allowing their translation into clinical epilepsy care and leading to future prospective clinical studies using HFOs to guide prospective clinical decisions.