While working on a “secret project” at Google Life Sciences (before it had been rebranded to Verily), Brian Pepin was struck with a question:
Why aren’t there analogous devices to glucose monitors for neurological diseases like Parkinson’s?
“At least in terms of the data, science, and symptoms, Parkinson’s is a more complex disease than diabetes, where you have nominally fewer kinds of inputs and outputs—you’re tracking blood glucose, or maybe you’re looking at A1C,” Pepin told Inside Precision Medicine. “But in Parkinson’s, you’re looking at multiple types of symptoms, including tremors and dyskinesia.”
That question inspired Pepin to leave the tech giant six years ago, start Rune Labs, and develop StrivePD—software that helps Parkinson’s patients track their experience and received FDA clearance to use an Apple Watch feature called the “movement disorder manager API” to record data about Parkinson’s tremors and dyskinesia automatically. In essence, StrivePD is a digital companion that keeps track of your Parkinson’s without having to be constantly thinking about the neurological condition.
But Pepin wants to go beyond tracking the data and making user-friendly reports. He wants StrivePD’s data to guide individualized treatment for Parkinson’s patients. That’s precisely what Rune Labs launched today with StrivePD-AI, which can take all of the data collected from StrivePD and write clinical reports to aid patients, caregivers, and clinicians in improving outcomes through optimized care.
Movement disorder specialist shortage
While Rune Labs’ technology is intriguing in its own right for its contribution to precision neurology, it is far more significant given the extremely long wait times for appointments with movement disorder specialists, which can last months. The combination of a neurology shortage and a sizable patient population of more than a million patients has resulted in wait times of several months in metropolitan areas where most movement disorder specialists work, and the wait times in rural areas are even longer. Traveling to the nearest movement disorder specialist in more remote locations may necessitate staying overnight at a hotel.
However, a single visit is typically insufficient for early-stage Parkinson’s disease when symptoms are often moderate, and a definitive diagnosis can be challenging to achieve. Later-stage patients need even more frequent visits. Medical care for Parkinson’s disease is extensive and long-term.
Pepin thinks Rune Labs can relieve pressure on health systems that depend on movement disorder specialists by offloading capacity for them. That is where the incorporation of generative artificial intelligence (GenAI) and large language models (LLMs) comes in, making everything StrivePD records more understandable and actionable.
Pepin said, “We were doing everything that we can to simplify visualizations of this data, but before the LLMs, the burden was on patients to interpret and pull up the most salient parts. That burden is far lower with GenAI involved. This multimodal data stream would be very hard or time-intensive for any human to analyze on a patient-by-patient basis. Access to these LLMs delivers personalized coaching and feedback and recommendations to patients based on this multimodal data stream.”
The GenAI algorithms developed by Rune integrate Anthropic’s Claude 3.5 Sonnet LLM—a rival to OpenAI’s ChatGPT—with all of the multimodal data collected by StrivePD. This data includes patient-reported outcomes like symptoms or medication side effects, activities tracked by Apple Fitness or logged into StrivePD, and free-form notes written by the patient, such as readings of blood pressure, sleep patterns, and mood changes. Utilizing Claude’s capabilities, the model compiles pertinent information from the most credible sources on Parkinson’s disease and generates individualized educational materials based on the patient’s StrivePD data. This information includes, but is not limited to, typical medication prescription practices and side effects, fall prevention strategies, and cognitive and sleeping symptoms.
Once a month, the patient receives an email from StrivePD-AI containing a comprehensive clinical report compliant with HIPAA regulations. To meet patients where they are, the report employs a motivational framework, encouraging them to keep recording their experiences and to notify their doctor of any balance or fall-related problems, which are considered high-risk occurrences. A thorough synopsis of notes, medication adherence, exercise records, and symptom fluctuations are all part of the clinical report.
At the end of the report, personalized educational materials are provided to patients. These include strategies to reduce the risk of falling, exercises to improve functional movements and mobility, guidelines for improving non-motor symptoms like mood and sleep, and exercises to promote large amplitude movements, faster movement initiation, and smooth transitions between rigidity and slowness.
“We help them close the loop on their own with their physiotherapist and then make sure that post-physiotherapy, we see those symptoms get better,” said Pepin. We can do all that, and the clinician can be informed but doesn’t have to take action, whether answering a secure email or a patient visit. The [movement disorder specialists] can then use that expanded capacity to see patients that are at higher risk, see new patients.”
Digital dopamine management system
Despite StrivePD-AI’s ability to analyze multimodal data streams in a way similar to that of a neurologist, Pepin insists that the technology is not intended to supplant human clinicians.
That being said, Pepin did mention that by thoroughly understanding the contents of a patient’s care plan, an approach like StrivePD-AI could look at what medications a doctor has approved and make suggestions for the medication schedule. For example, Pepin said that perhaps insights could be used to move the timing or regimen of levodopa.
“We’ve evolved the core product into something that’s starting to look like a ‘dopamine management system’ and moving into better models for how Parkinson’s progresses in different patients,” said Pepin. “We have a collection of things that are low-risk routine and within a patient’s care plan. We’re not writing new scripts; the clinician is informed of what happens.”
Rune Labs has also entered the business of developing disease-modifying therapies, albeit indirectly. While Rune Labs does not have a drug discovery program or pipeline, Pepin believes that StrivePD-AI will help companies with whom we collaborate design and run their trials in radically more efficient ways, allowing them to run at a lower cost, faster, and with a higher likelihood of success.
Over the last few years, Rune Labs has announced several collaborations, including one with Aspen Neuroscience to remotely collect patient data for clinical trials testing autologous cell therapy for Parkinson’s. In March 2023, Rune Labs and BlueRock Therapeutics, a biopharmaceutical company in the clinical stage and a wholly-owned subsidiary of Bayer AG, announced a partnership. The partnership is centered around StriveStudy, Rune Labs’ clinical trial platform, which includes data collection and patient monitoring tools. The goal is to create a comprehensive and accurate picture of the activity of Parkinson’s disease and to improve patient engagement in the study.
Recently, Rune started working with health plans like Kaiser Permanente to coordinate Parkinson’s care for patients with the help of their current clinical team. LLMs offer services like medication adherence and coaching. Pepin said that Rune has developed a version of their technology that is deployed specifically with Kaiser, an approach he’d like to see applied to more health systems. To make this available to more people, Pepin is working to establish partnerships with health systems outside of Kaiser and with insurance companies.
The cherry on top for Rune Labs would be to turn their technology into a digital diagnostic. Pepin said, “My feeling is that if any disease is diagnosable via wearable data, it’s going to be Parkinson’s, and partially that has to do with the way Parkinson’s is defined. Now that we have this in place, we’re experimenting around and showing that we’re getting actual clinical results, like keeping people out of emergency rooms. Healthcare is just slow to change; it’s a long journey. We’re six years in, and sometimes I feel like we’re just getting started.”
What’s preventing widespread use?
Compared to the efforts of digital pathologists to promote whole slide imaging among pathologists, Rune Labs’ mission is far more difficult. Aside from convincing doctors and health systems to use Rune’s iOS illness management app, there is one additional stumbling block: until another activity-tracking device receives FDA approval, patients must buy an Apple Watch.
Pepin and Rune Labs are not the only contenders in this race; there is some competition for the best app that uses iOS and Apple Watch. The FDA granted H2o Therapeutics and its Parky app the same clearance as StrivePD for the movement disorder API. Then there is the Android side, where participants in the Parkinson’s Progression Markers Initiative (PPMI) study can access the Roche PD Mobile Application. However, because the Roche PD Mobile Application only uses an Android phone and no accompanying wearable, the passive data collected will likely be less advanced than its iOS counterparts.
It is unclear whether Rune Labs is making any headway, let alone investing in Android or Fitbit. Rune Labs could be secretly developing other OSes and wearables. After all, it just takes a few minutes for an LLM like ChatGPT or Claude to spin out a fitness app blueprint, complete with the code to access Android sensors like accelerometers.
So, maybe more important than developing new features for StrivePD is determining how to make Rune Labs’ approach available to the public. That might mean more OS and wearable compatibility in the U.S. What about the rest of the world? That could necessitate a novel, inexpensive technology that could be mass-produced anywhere, like Openwater’s open-source systems based on cell phone chips. That is, of course, assuming that mobility disorder specialists suddenly appear worldwide, let alone in the United States.
Maybe one day, there will be no need for patient care at all, given the recent data on cell therapies for Parkinson’s disease, like the positive results for bemdaneprocel’s ability to restore dopamine 18 months after treatment! Until then, Pepin and Rune Labs have their work cut out to improve Parkinson’s disease care.
