Skip to content
The Story of Rune Labs
Neurotech, Venture Portfolio

The nervous system is the substrate of human experience. Logic would therefore predict, and experience would confirm, that some of the most phenomenologically pernicious diseases and disorders are those that implicate the nervous system. This specific insight is the motivation behind our commitment to neurotechnology at Loup, and it’s within this context that I’m extremely excited to share publicly our newest neurotechnology investment: Rune Labs!

Rune, under the direction of CEO Brian Pepin, is developing a software platform for brain data.

Laying out some context

If there’s one high-level trend I’ve identified in neuroscience and neurotechnology—via literature and via conversation–it’s that circuits are in vogue. The central and peripheral nervous systems are structures of dumbfounding complexity and faced with dumbfounding complexity, the classic human recourse is to abstract, abstract, abstract until we’ve reached a level of granularity that lends itself to comprehensible analysis and scientific investigation. A brain can be investigated at the level of a molecule, such as when studying the structure of a membrane protein. A brain can be investigated at the level of statistical characterization of a large and stochastically-interacting set of molecules, such as when studying ion fluxes and action potentials. A brain can be investigated at the level of microbiology, such as when studying cell migration during neural development. And, lastly and importantly, a brain can be investigated at the level of neural circuits. It’s this last level of abstraction that’s seen strong uptake in recent years, in both research and translational settings. 

The neurocircuitry level of abstraction is prominent in basic science, but it also shows its face in clinical research: in the translational setting, the neurocircuitry view has yielded the field of neuromodulation. Neuromodulation is by no means new, but it’s growing in size and publicity. A non-exhaustive list of modulation technologies that leverage the circuitry approach includes vagus nerve stimulation, transcranial direct/alternating current stimulation, transcranial magnetic stimulation, and deep brain stimulation (DBS). Here, we’ll focus on DBS. DBS has FDA approval for use in Parkinson’s Disease (PD), and to date, more than 160,000 patients have been implanted for PD and other indications. “Other indications” refers to the investigative work happening in, for example, treatment-resistant major depressive disorder, treatment-resistant obsessive-compulsive disorder, and addiction disorders (among others).

Original DBS devices—i.e., most of the devices currently implanted in commercial settings—only have the capability to stimulate the brain. This makes it challenging to customize therapy to individual patients, and impossible to design therapies that use realtime information about circuit activity to determine when/how to stimulate. A new wave of devices is under development that can both record and stimulate; these are referred to as “closed-loop” DBS devices. 

Confluence of problems

Brian founded Rune to solve a confluence of problems. At the most abstract level, brain disease is complex because brains are complex; this means that any data recorded from the brain will be hard to analyze. Furthermore, in the domain of psychiatric disorders wherein the disorders are characterized largely by their qualitative, phenomenological symptomatology, there’s the added challenge of finding ways to measure phenomenological states—which is arguably impossible in a philosophical sense.

There are also more concrete problems facing the treatment of neurological and psychiatric disorders via neurotechnologies such as DBS. 

  • Noise. Any brain measurements, whether electrical or otherwise, will contain some degree of noise. Neurophysiological data is notoriously noisy, meaning certain recording segments can’t be used since they’ll skew analyses.
  • Longitudinality. PD, TRD, etc. all progress and/or fluctuate over long periods of time. PD, for example, is known to have a circadian fluctuation, which is notoriously difficult to capture through discrete clinical visits. Tracking these diseases/disorders must therefore require researchers to capture data over a long time span; this impacts clinical trial design, algorithm/treatment development, and therapy delivery in commercially-implanted patients. In addition to the idea of longitudinality having paradigmatic significance, there are practical concerns: large data sets are hard to work with. Some things that can contribute to the size of electrophysiological data are, for example, the number of recording channels, the sampling rate, and the duration of the recording. Closed-loop DBS, in particular, will create longitudinal electrophysiological recordings that haven’t existed before: with RC+S being implanted and the Percept trial on the horizon, there is likely to be more invasive human neural data collected in 2020 than every previous year in history combined.
  • Patient environment. The patient’s environment interacts with brain disease, and therefore environmental factors have to be considered and possibly tracked when developing and delivering therapies.
  • Patient experience. If neural data comprises “measurements,” such as the data recorded by a closed-loop DBS system, then in order for a supervised learning algorithm to know how to respond to the measurements, it must decide how those measurements relate to positive patient experience vs. negative patient experience. This amounts to collecting labeled neural data from patients, and infrastructure to do so is generally poor.
  • Complex feature spaces. Given a time series of neural data (or any other kind of neural data), there are many possible ways to transform that data to derive useful “features” for use by a learning or control algorithm. Handling large feature spaces, as well as selecting high-performing features, requires time (and therefore money) as well as computational power. 

Each of these categories can and do fill books. The conclusion, though, is simple: treatment of neurological and psychiatric disorders has miles and miles of road left to build and travel, and many of these miles will be paved by data.

Time series neural data (electrophysiology and/or fMRI) is so dense and complex that it isn’t human-readable. Furthermore, any model we have that goes beyond more than a few neurons interacting is necessarily a massive simplification and surely throws out many mechanisms. Brian believes these factors motivate the need to use modern deep learning approaches to understand the data and generate new insights. However, to pull this off requires a lot of data that is harmonized at some level by similar labels, sampling characteristics, etc. Data is siloed right now, sometimes even at the level of a single patient clinical visit, preventing harmonization—and that is where Rune comes in: Rune provides unified software interfaces to a large number of patients, researchers, and clinicians that solve the issues outlined above and allow the aggregation of datasets amenable to AI.

Rune’s solution

Rune is developing a software platform for the labeling, ingestion, cleaning, visualization, and analysis of intracranial brain data. In the technology world, data has scaled, and therefore so has the intricacy of the software engineering necessary to derive value from the data. Brian and his team are bringing the same strategy to neurotechnology.

In conjunction with a researcher-facing platform, Rune has launched a patient-facing application for Parkinson’s patients to track symptoms; this data is fused with neural recordings on the backend to give physicians and researchers better insight.

Rune’s software platform moves the needle on the problems we discussed above:

  • Noise. Rune’s software automatically removes chunks of data that are too noisy to use.
  • Longitudinality. The Rune platform is being engineered to handle longitudinal data efficiently; this will expedite R&D of therapies that incorporate longitudinal changes/trends in patient behavior and neural data. 
  • Patient environment. Through the patient app, Rune can collect and supports the analysis of data corresponding to environmental variables of interest.
  • Patient experience. The patient-facing application is designed from the ground up to capture patients’ experience with their conditions and to use that data to understand the impact of neurotechnological interventions.
  • Complex feature spaces. Rune has various efficiently-implemented data transformations and visualizations built into the platform, enabling researchers to quickly investigate data and test hypotheses.

Where they are and where they’re going

Near-term

Today, Rune launched their patient-facing app for PD patients, called Strive PD. The app was designed by people with Parkinson’s, for people with Parkinson’s. It serves as a patient portal to information about their therapy and lets patients record relevant information such as their symptoms and whether or not they took their medication. If patients want, they can share the data with clinicians and caregivers to help find the right mix of therapy. In the context of new therapy development, Rune can adapt its baseline patient app to support new use-cases, including direct wireless connection to implanted devices.

Rune is also actively partnering with global leaders in the development of DBS therapies. Initially, they’re focusing on accelerating development of improved DBS therapies for Parkinson’s, and as time goes on they’ll expand to other indications.

Long-term

As with most things in our world these days, the more patients we help, the more patients we can help; in other words, we use the data from present patients to help future patients, and this cycle compounds exponentially as the datasets grow. It seems likely that psychiatric and neurological healthcare continue on their data-centric trajectories, which means there will be a tension between the convenience of non-invasive and/or proxy data—such as data collected from wearables, smartphones, and ambient sensors—and the “closeness to the source” of intracranial brain data. Rune’s long-term vision is to expedite and validate the development and commercialization of new neurological and psychiatric therapies by partnering with other researchers and businesses: as therapies get more complex and personalized, the primary innovation will be in the development of new software and algorithms—right at the nexus of Rune’s expertise.

Conclusion

If one thing is certain about the direction of brain science, it’s that the quantity and precision of data we have about the brain is only going to increase. Therefore, there is a strong need to leverage that data to develop therapies, and we’re thrilled to be working with Brian and his team at Rune to do just that.

Disclaimer

Back To Top