In 1817, James Parkinson first described Parkinson’s Disease as a distinct neurological syndrome.  Two-hundred four years later, we’ve made a lot of progress in diagnosis and measurement of symptoms, in identifying risk factors, and in treating the most prevalent motor symptoms.  However, we still don’t know what Parkinson’s is: it has been challenging to link clinical symptoms to potential underlying disease mechanisms, and this has been the major barrier to developing disease modifying therapies.  Unlike oncology, where data collected from patients can be used to develop and deliver highly precise therapies, “precision neurology” care is currently blocked by a lack of access to useful data about a patient’s specific disease. Parkinson’s researchers working to develop better therapies wish for better access to high-quality human data about the brain, but historically this data has been siloed and limited in scope.

Today, I’m excited to share that Rune Labs’ collaborators at UCSF have published groundbreaking research in Nature which shows how newly available clinical brain data is changing our ability to understand and treat brain diseases like Parkinson's.  The publication summarizes insights from the first 2500 hours recorded of high-bandwidth Parkinson’s brain-sensing data in patients with implanted Deep Brain Stimulation (DBS) devices, and gives a new window into the brain function of people with Parkinson’s that can help bridge the current gap between biology and clinical practice.  The big implication here is that direct brain sensing from neural circuits implicated in Parkinson’s is now possible *in humans* at scale.  This is possible because DBS devices - used for two decades as a treatment for motor symptoms in Parkinson’s -  are now incorporating brain sensing in addition to therapeutic stimulation.  With ~20,000 DBS devices implanted annually, this represents a multiple-orders-of-magnitude explosion in the availability of high-quality electrophysiology data about Parkinson’s disease, and is already leading to new insights (see below).

Unlike previous research which was limited to minutes (or at most hours) of in-clinic data, the UCSF team used a newer-generation device which could record brain sensing data at-home as patients go about their normal daily life.  This is a big deal because Parkinson’s, like most neurological conditions, can vary significantly over the day, day-to-day, in response to different environmental pressures, and in response to medication and diet: any in-clinic snapshot will only give a limited picture of a patient’s overall disease phenotype.  By contrast, this publication shows 2500 hours of brain sensing over days, weeks, and months across 5 individual patients, including quantitative tremor and dyskinesia labels from wearable devices and patient-reported outcomes for medication and symptoms.  Lead researcher on the paper Ro’ee Gilron aptly described this as a “Hubble Telescope” moment for PD research in the press release that accompanied the publication. Being able to continuously sense invasive brain signals at this scale - compared to what has previously been possible - is like using a space-based telescope vs a ground-based telescope.  You can see more clearly, you can resolve entirely undiscovered phenomena, and it's going to have a huge and lasting impact as we figure out better and better how to use it (again, just look at the data revolution in oncology over the past two decades, with a corresponding drop in mortality of >30%).

As researchers first turned on this “Hubble Telescope” for the brain, what did they find in the initial data?  Two things stand out:

1. Patient-specific neurological fingerprints - By observing neural signals evolving over the course of entire circadian cycles, and also seeing repeated cycles over the course of many months, the researchers were able to study phenomena like oscillatory activity during on/off transitions and Parkinson’s sleep pathology at a level of detail not previously possible.  While some of these effects have been observed in short-term recordings in clinical environments, the much larger amount of data available in this study allowed for within-individual and showed that the way these signals vary over time may constitute a kind of neurological “fingerprint”. In the paper, the researchers leveraged these “fingerprints” to deliver optimal DBS therapy for each patient, but there is a clear opportunity to extend this to optimizing therapy holistically, including for pharmaceuticals.

2. Electrophysiological disease phenotypes for Parkinson’s -  Leveraging some of the neurological fingerprint work described above the researchers show evidence of at least one distinctive electrophysiological phenotype that corresponds to a distinctive clinical phenotype.  More patients are needed as the sample size here is small, but if confirmed this would be one of the first viable pathways to connect a clinical phenotype back to an underlying disease mechanism (protein agglomeration, genetic mutation, etc).  Also, it stands to reason that if there is evidence of one phenotype already, studying more patients will probably reveal more phenotypes.  Better understanding of Parkinson’s phenotypes will lead to more efficient clinical trials, including for disease modifying drugs, and better understanding of new therapeutic targets.

Of course, this is just the beginning: 5 patients, and 2500 hours brain signal recording.  Since the data for this paper was collected, we’ve partnered with researchers at UCSF and around the world to scale up data labeling, ingestion and aggregation of large-scale brain sensing data.  Most of this work has occurred under various forms of COVID lockdown, which is possible because brain recording from these devices – and the wearable devices like Apple Watch that can be used to generate important labels – can be done remotely.   We are eager to see what new insights are revealed by the first 25,000 hours of recording, the first 250,000 hours of recording, and beyond, and to translate these insights into better therapies for Parkinson’s.