Bexorg: Transforming CNS Drug Discovery with Whole-Brain AI

Most attempts to treat brain disease start with the same problem: we don’t really know what happens inside a living human brain.

For decades, drug development in neurology has depended on proxies—cells in dishes, animal models, and simulations that can’t capture the complexity of human neural networks. The results speak for themselves. More than 95 percent of central nervous system (CNS) drugs fail before reaching approval. Alzheimer’s disease went nearly two decades without a single new therapy.

A startup out of Yale, Bexorg, believes it can change that. The company has built the first integrated AI and whole-human brain platform for drug discovery, blending wet-lab experimentation with machine learning to study living human brain tissue in unprecedented detail.

Founded by Yale neuroscientists Dr. Zvonimir Vrselja and Dr. Nenad Sestan, Bexorg emerged from the same research that first stunned the scientific community in 2019—work showing that metabolic and molecular activity could be partially restored in postmortem brains. The company’s mission now is to turn that discovery into infrastructure for drug development.

If it succeeds, the implications are enormous: a system that lets researchers test compounds directly in metabolically active human brains before clinical trials begin.

The Crisis in CNS Drug Development

Few areas of medicine are as unforgiving as neuroscience. The human brain is vastly more complex than any model we’ve built to study it, and most drugs that look promising in cells or animals fall apart in people. Translational failure is the rule, not the exception.

The problem isn’t just scientific—it’s technological. Traditional preclinical research depends on rodent models, cell cultures, and organoids that lack the full architecture of a functioning brain. Microfluidic systems and brain-on-a-chip devices have helped bridge the gap, but they can’t reproduce the organ’s true physiology: its cellular diversity, electrical activity, or regional interactions.

That’s the gap Bexorg wants to close. Its BrainEx platform uses a perfusion system to restore metabolic and molecular activity in intact human and pig brains, keeping tissue alive long enough to test real drug responses. The system circulates a custom artificial blood solution through a network of tubes that mimic the body’s own vasculature, allowing compounds to reach neurons, glia, and vascular cells much as they would in life.

The result isn’t a sentient brain, but a functional one—an organ that still processes energy, maintains structure, and exhibits cell-specific behavior. It’s a way to study the human brain as an active biological system, not just a collection of slices under a microscope.

Platform Technology & Capabilities

At the center of Bexorg’s approach is BrainEx, a custom-built perfusion system that looks more like a clinical instrument than a lab prototype. Rows of transparent reservoirs feed a network of pumps, filters, and tubing that continuously circulate an oxygen-rich solution through whole, donated human and pig brains. The solution mimics the composition of blood, carrying oxygen and nutrients while clearing waste, allowing the organ’s cells to resume limited metabolic and molecular activity.

This process doesn’t restore consciousness or higher brain functions, but it does revive the tissue’s physiology—neurons fire, glial cells regulate the chemical environment, and microvasculature responds to flow. For researchers, that means the ability to study drug interactions in a living, intact brain rather than in dissociated cells or thin tissue sections.

The setup is also compatible with the same analytical tools used in modern neurobiology. Samples from perfused brains can be examined by RNA sequencing, mass spectrometry, or imaging mass cytometry to capture transcriptomic, proteomic, and metabolic profiles. Microdialysis probes measure neurotransmitter levels in real time. These layers of data create a detailed map of how drugs move through and act within human brain tissue—something even the best animal models rarely get right.

Bexorg’s facility in New Haven currently runs five BrainEx machines, with a sixth under construction. The company sources both healthy and diseased brains through established organ donation networks, enabling side-by-side comparisons of how compounds behave in different neurological states. Each experiment generates a dataset capturing everything from synaptic signaling changes to metabolite flux, forming what’s becoming one of the most complete human CNS datasets in existence.

The scale is growing fast. Bexorg plans to run 1,000 whole-brain experiments per year by next year, spanning Alzheimer’s, Parkinson’s, and other neurodegenerative conditions. For drug developers, that represents a preclinical testing environment that finally reflects the human biology their drugs are meant to target.

XO Digital: Closing the AI Loop

BrainEx produces something researchers have always wanted but rarely had—high-fidelity, human-based experimental data. Bexorg’s AI platform, XO Digital, turns that data into insight.

Each BrainEx experiment feeds directly into machine learning models that analyze molecular signatures, drug-response patterns, and biomarker changes. Over time, these models learn to predict how compounds will behave before they’re even tested, creating a feedback loop between wet-lab validation and computational prediction.

The system runs on petabyte-scale datasets, including transcriptomic and proteomic measurements from hundreds of whole-brain experiments. Unlike most AI systems in neuroscience, which rely on public or simulated data, XO Digital is trained exclusively on experimentally generated human data. This gives it a level of biological grounding that’s almost unheard of in drug discovery.

In practice, the platform works like an autonomous discovery assistant. It can flag promising drug targets, suggest new biomarkers, or identify safety risks early in development. The more data the lab produces, the smarter the AI becomes—tightening the loop between experiment and prediction until each informs the other in real time.

What emerges is a new model for CNS drug discovery: one that starts in human tissue, learns from it, and keeps learning as the data accumulates. For a field defined by uncertainty, it’s a rare glimpse of predictability.

Leadership & Scientific Foundation

Bexorg’s origins are unusually direct. The company didn’t start from a business plan—it started from a discovery that forced neuroscience to reconsider its assumptions.

In 2019, Yale researchers Dr. Zvonimir Vrselja and Dr. Nenad Sestan published a study in Nature that sent shockwaves through the field. Their team had restored cellular and metabolic activity to postmortem pig brains hours after death, using a perfusion system that became the prototype for what is now BrainEx. The work was part of the NIH’s Brain Initiative, a broader effort to map and understand the human brain at every level.

The experiment didn’t bring the brains “back to life,” but it proved something that had been unthinkable: brain cells, even in intact tissue, could resume function well after circulation had stopped. For drug discovery, that meant researchers could finally observe how compounds interact with intact human brain circuits, not just cultured cells or animal analogs.

Today, Vrselja serves as Bexorg’s CEO and co-founder, guiding the company’s transition from research breakthrough to operational biotech. He combines medical training with a deep understanding of the technology’s ethical and scientific boundaries. Dr. Sestan, still active at Yale, continues to advise the company on scientific direction and model development.

Bexorg has grown from a small academic spinoff into a multidisciplinary team of 30 scientists and engineers, spanning neuroscience, data science, and machine learning. Drug discovery operations are led by Paul Wes, a former Pfizer executive who brings experience managing translational programs across neurology and psychiatry. Together, the team is building something that sits between academia and industry—a research organization engineered for scale.

Funding & Strategic Growth

Bexorg has raised $42.5 million to date, a modest figure by AI-biotech standards but one that signals focused execution. Its $23 million Series A, announced earlier this year, was led by Engine Ventures with participation from Connecticut Innovations, E1 Ventures, Amplify Partners, and Starbloom Capital.

The round added Dr. Ann DeWitt, General Partner at Engine Ventures, and David Beyer, Partner at Amplify Partners, to Bexorg’s board. DeWitt called the company’s platform “redefining CNS drug discovery,” emphasizing its ability to shorten development timelines and improve clinical success rates.

The capital is being used to scale both sides of the business: the BrainEx wet-lab infrastructure and the XO Digital AI engine trained on human brain data. More machines are being built, more data is being generated, and new disease programs are being seeded with partners.

The financing also reflects growing investor appetite for “infrastructure science”—platforms that sit upstream of traditional drug development. Bexorg isn’t chasing a single therapy. It’s building the tools that could make hundreds of therapies more likely to work.

Strategic Partnerships & Revenue Model

Bexorg isn’t trying to build the next blockbuster drug on its own. Its strategy is to become essential infrastructure for others who are.

The company partners with pharmaceutical firms and academic institutions, providing access to its BrainEx platform and XO Digital engine as a way to evaluate compounds directly in metabolically active human brain tissue. For drug developers, that access can reveal whether a molecule reaches its target, how it behaves once it does, and whether the early biology holds up before a single clinical trial begins.

In mid-2025, Bexorg signed a multi-program collaboration with Biohaven Ltd. (NYSE: BHVN)—a company well known for pushing the boundaries of neuroscience R&D. Biohaven has already tested two compounds on the BrainEx system: one focused on boosting brain metabolism and another targeting Parkinson’s disease. According to CEO Vlad Coric, the experiments provided enough insight to fast-track both programs. “It’s like fast-forwarding the research timeline,” he said. “You can see early whether you’re heading in the right direction.”

Earlier, Bexorg joined forces with the University of Oxford and the UK Medical Research Council to support translational gene therapy research. The goal there is to evaluate gene-based therapies in real human brain tissue—something traditional preclinical models can’t do effectively.

Beyond partnerships, Bexorg is building a data business. Each experiment expands a proprietary repository of human CNS data that can be mined for new targets, biomarkers, and molecular patterns of disease. Pharmaceutical companies can license access to these datasets or partner with Bexorg to run targeted discovery programs. The model blends service revenue with long-term data compounding: the more experiments the company runs, the more valuable the platform becomes.

Market Opportunity & Impact

The need for better brain models isn’t theoretical. More than 70 million people live with neurological diseases such as Alzheimer’s, Parkinson’s, and multiple sclerosis. Each represents billions of dollars in research and development spending—much of it wasted when promising therapies fail in human trials.

That failure isn’t just about biology. It’s about translation. Animal models don’t mimic the human brain’s circuitry or molecular nuance. Even the most sophisticated human cell cultures lack the structural and vascular context that determines how drugs behave in real tissue.

Bexorg’s platform fills that void. By creating living, perfused human brains that can be probed and measured with tools like transcriptomics, mass spectrometry, and multiphoton imaging, the company gives drug developers an early, human-relevant readout of efficacy and safety. It’s a shift from discovery by proxy to discovery in the real thing.

For pharmaceutical partners, the payoff is speed and confidence. A drug that would take years to reach an early clinical signal can now be triaged in months. Those insights can save millions of dollars and prevent entire programs from collapsing late in development.

That’s why investors see Bexorg not as a single-asset biotech, but as an emerging layer of research infrastructure. If successful, its platform could underpin much of the next generation of CNS drug discovery, transforming what’s long been one of medicine’s most intractable fields.

Looking Ahead

Bexorg sits at an unusual crossroads: part neuroscience company, part technology platform, part ethical experiment. The work pushes scientific boundaries while demanding constant moral reflection.

The company’s approach—using donated postmortem human brains to restore limited molecular activity—exists in a delicate space between research and respect. As CEO Dr. Zvonimir Vrselja has explained, the goal is to “restore specific molecular activities essential for drug discovery, while higher brain functions are not restored.” The distinction matters. Bexorg operates under rigorous bioethics standards and oversight by an independent ethics board. Each donated organ is treated not as a resource, but as a gift, used to generate knowledge that could eventually help millions.

The science, however, speaks to something even larger: how biology and computation are merging into one continuous system. By pairing wet-lab experimentation with AI that learns from every result, Bexorg is constructing a model of discovery that could eventually become self-improving—a system where each brain experiment refines the next, and the AI grows increasingly adept at predicting outcomes before they happen.

If it works, the implications go far beyond CNS drug discovery. The same framework could be adapted to cardiac, hepatic, or renal tissues, allowing researchers to probe diseases organ by organ, using human biology as the model.

The immediate focus remains on the brain, where unmet need is highest and risk greatest. With fresh funding, partnerships spanning academia and industry, and a platform that merges ethics, AI, and physiology, Bexorg is defining a new category of research infrastructure: whole-human experimentation for the age of computation.

The question is no longer whether we can study the human brain more directly. It’s whether we can handle what happens when we finally do.