Using Big Data to Personalize Epilepsy Care

December 17, 2018

The professor and head of the Department of Neurology and Rehabilitation at the University of Illinois at Chicago spoke about how data can better inform epilepsy care.

Jeffrey Loeb, MD, PhD

Big data sets have found their way into many specialties in medicine, fitting nicely with the ongoing push toward personalized care, but where they’ve become even more helpful is in conditions which are complex and often misunderstood.

Epilepsy, according to Jeffrey Loeb, MD, PhD, professor and head of the Department of Neurology and Rehabilitation at the University of Illinois at Chicago, is one of those conditions. Although there is much still left to be parsed out, the medical understanding of epilepsy continues to grow over time with the use of large data sets.

At the Child Neurology Society’s annual meeting, Loeb gave a presentation detailing how much can be gleaned by using big data. In a conversation with NeurologyLive he provided some insight into how it has helped change the landscape in epilepsy.

NeurologyLive: You gave a presentation at the Child Neurology Society’s annual meeting about utilizing big data in epilepsy. Could you share the major takeaways from that?

Jeffrey Loeb, MD, PhD: Epilepsy is probably one of the most complex diseases that we understand very little about, and sometimes it’s hard to know where even to begin. The personalized approach that we take is to learn from our patients who have epilepsy, and by studying those patients that we probably know more about than just about any other disease—patients who undergo evaluation for epilepsy surgery.

Prior to surgery, we have to make sure we remove the epileptic parts and keep the good parts of the brain still intact. To do that, we do a lot of imaging studies—brain imaging studies such as magnetic resonance imaging (MRI) and positron emission tomography (PET) scans. We do electrical recordings, not only on the scalp, but often intracranially where we remove the scalp from the skull and put hundreds of electrodes to identify the areas that are epileptic and the that aren’t. Then, when tissue is removed, we’ve taken the unique approach of taking tissue of each of those locations that has different types of epileptic activity, mapping those networks in those patient’s brains, and creating a 3D model. That incorporates the brain, the imaging, and everything we know about the patient from the electronic health record, to the electrical recordings, to the tissue, to the histology from the tissue, the genes, the proteins, the molecules.

Really, this is a big data approach, but by studying individuals who have epilepsy. Once we’ve done that, we look for common features in most patients who have epilepsies with goals of generating new therapeutics, new biomarkers, and new ways to diagnose and treat epilepsy.

What do we know now that we didn’t know in the past 5 to 10 years?

The high-throughput approach that we take by looking at every gene in a piece of tissue—we don’t have a hypothesis other than there’s a part of the brain that does something different that produces seizures. So, what’s unique about that? Based on the constellation of genes, proteins, and molecules, we’ve discovered that there are certain layers of the cortex that are activated and different for interictal spikes versus seizures. We then use animal models to identify some of these pathways that we discovered are actually necessary and sufficient for seizures, and then we develop drugs against those pathways.

We have learned that certain lesions—we call these ‘micro-lesions’—in the brain that we never knew before, and now we know how to stain for them based on the constellation of genes and markers that we discovered in the brain.

The other thing that we’ve learned how to do is map the epileptic activity better. Many people look at seizures and they look at spikes, and when they look at the spikes, which are very frequent between seizures, we just count them—how many at this location versus that? But spikes are not sedentary, they actually migrate around the brain. So, we’ve developed some approaches to map the migration of spikes and look at their network pattern and relate those now to seizure network patterns and get a better understanding of the brain networks. We move all the way down through the imaging, to the electrical networks, to the genes, to the proteins, to the molecules with a goal of generating a personalized approach—not just for one patient but many patients who have epilepsy.

Looking to the future, what’s your hope? What question are still left unanswered?

There are still tons of questions we have left to answer. We’re still trying to sort out what’s different and what the relationship between the spikes and the seizures are and what are the genes and pathways responsible for each. We ask whether there ways we can treat the disease without surgery someday—either through small molecules or devices—so that by understanding the networks we can do a better job.

The other thing we’re excited about is recently discovered metabolites that are unique to epileptic brain regions. In addition to proteins and genes, we look at small molecules called metabolites and we used the same NMR that we used for MRI studies to identify a unique signature for these metabolites. What we’re hoping to do is develop a way to diagnose and treat epilepsy with the same precision we get with intracranial recordings using an MRI scanner. If successful, we will be able to diagnose it earlier and develop treatments and use these markers as a predictive biomarker of future treatments.

Are there any misconceptions or stigmas that need to be addressed?

Well I mean “personalized medicine” is kind of a sexy term these days, so that we can look at the genetic background of someone and predict which treatment is good for one or another patient, and it’s a big challenge. We often look at 1 or 2 genes when we do these personalized medicine studies, and the problem that we’re finding is a lot of these pathways in epilepsy involve hundreds, if not thousands, of genes. So, to say that from a couple of genes we can make accurate predictions of who’s going to respond to which medication may be harder than it seems.

Plus, I don’t think we really understand the actions of a lot of the drugs that we say we currently use, so saying that a sodium channel drug is really a sodium channel drug may not be true. It may work on calcium channels or it may work on GABA receptors, or it may work completely differently for all we know.

Transcript edited for clarity.