AI's 'Virtual Biopsy' Aims to Solve Neurology's Billion-Dollar Problem

SAN FRANCISCO, CA – June 16, 2026 – In a move that sends ripples across the pharmaceutical and tech industries, the AI frontier lab Verge Labs today unveiled a new foundation model, vBx-1.0, that it claims can create a “virtual biopsy” of the human brain from a simple blood draw. The announcement tackles one of the most intractable and expensive challenges in modern medicine: understanding and treating neurological disease.

For decades, the brain has remained a biological black box. Unlike cancer, where a surgeon can physically sample a tumor to guide treatment, the living brain is inaccessible. This fundamental hurdle is a primary reason why an estimated 94% of drugs for neurological conditions fail in clinical trials, often after hundreds of millions of dollars have been spent. Verge Labs, a firm backed by a formidable roster including Eli Lilly, BlackRock, and Y Combinator, asserts its technology can finally bring the precision medicine revolution that transformed oncology into the realm of neurology.

The Digital Window into the Brain

At the heart of the announcement is vBx-1.0, an AI model designed to reconstruct the molecular state of a patient’s brain with what the company claims is up to three times the accuracy of previous state-of-the-art methods. By analyzing biomarkers in the blood, the model infers gene expression activity within the brain, effectively creating a detailed biological snapshot without invasive procedures.

This capability extends beyond mere diagnostics. Verge demonstrated the model's predictive power using data from a large Parkinson's Disease patient cohort. The results suggest vBx-1.0 can predict a patient's response to levodopa, the standard treatment, more accurately than existing clinical measures. The company estimates this could enrich clinical trials with 33% more likely responders, potentially shrinking trial sizes by a staggering 43%.

Further validation came from Verge's own Phase 1b trial for an ALS drug, VRG50635. In that trial, about a third of patients discontinued the drug after one dose for unknown reasons. When vBx-1.0 was applied retrospectively to baseline blood samples, it identified that the biological pathway targeted by the drug was already suppressed in the very patients who later showed intolerance. This suggests the model could have prospectively screened out approximately 34% of these patients, saving time, money, and sparing individuals from an ineffective treatment.

Recalculating the Economics of Drug Discovery

The implications of this technology are profoundly economic. The average cost to bring a new drug to market hovers around $2.8 billion, a figure inflated by the high cost of failed trials. In neurology, these costs are amplified by slow patient recruitment and the sheer complexity of the diseases. Delays in clinical trials can cost sponsors anywhere from $600,000 to $8 million per day.

Verge’s vBx-1.0 is positioned as a powerful de-risking tool. By identifying the right patients for the right drug before a trial even begins, it offers a path to smaller, faster, and cheaper clinical studies. “If you have a trial that failed in a mixed population, or you're designing one now, we want to run the virtual biopsy on it,” said Alice Zhang, Verge’s Chief Executive, in a statement that is as much a business pitch as it is a scientific call to action.

This value proposition has not been lost on major pharmaceutical players. Verge Labs already boasts collaborations with Eli Lilly and AstraZeneca's rare disease unit, Alexion, totaling a reported $1.6 billion in potential milestones. The company's new platform-first commercial model, exemplified by a recent CNS partnership with Tenacia Biotechnology, signals a strategic shift. Rather than just developing its own drugs, Verge is positioning itself as an essential data infrastructure and intelligence layer for the entire industry, a move underscored by its recent rebranding from Verge Genomics to Verge Labs.

Data as the New Bedrock of Biology

Verge’s competitive moat is not just its algorithm, but the decade-long effort to build its proprietary data corpus, VergeDB. The vBx-1.0 model was trained on a massive dataset comprising more than 12,000 brain transcriptomes from 6,500 patients, alongside matched genomic, proteomic, and clinical data. Crucially, this includes a physical inventory of over 900 frozen human brain tissue samples.

This “all-in-human” approach stands in stark contrast to traditional methods that rely heavily on animal models, which often fail to predict human responses. As CEO Alice Zhang vividly put it, “We increasingly think of brain tissue as ‘LiDAR for neuroscience.’ Without it, every blood or imaging proxy is a flat picture, missing depth.” This investment in deep, multimodal human data is what allows the model to bridge the gap between an accessible peripheral signal (blood) and the complex, inaccessible biology of the brain.

This strategy is reflective of a broader trend where proprietary, high-quality data has become the most valuable asset in the AI arms race. By creating a feedback loop where data from its own clinical trials is used to refine its models, Verge is building a continuously learning system that becomes more powerful and accurate over time.

A Crowded Field Navigates New Rules

Verge Labs is not operating in a vacuum. The promise of applying AI to drug discovery has attracted immense talent and capital, with a growing number of companies like Insilico Medicine, BenevolentAI, and Karavela AI developing their own platforms. Some focus on generating novel drug candidates, while others, like NeuroDiscovery AI, specialize in optimizing clinical trial recruitment.

What distinguishes Verge is its deep focus on human brain tissue as the ground truth for its models. However, as this technology advances toward clinical application, it enters a complex regulatory and ethical landscape. Regulators are moving to keep pace; the FDA recently issued updated draft guidance on the use of AI in drug development, establishing a risk-based framework for model validation.

Furthermore, the use of AI trained on vast patient datasets raises critical questions about data privacy, security, and algorithmic bias. Ensuring that models are trained on diverse and representative populations will be essential to prevent exacerbating existing health disparities. Building trust with patients and the public will require a high degree of transparency about how these powerful “virtual biopsy” tools work and how patient data is protected. For Verge Labs and its competitors, navigating these technical, commercial, and ethical challenges will be as critical as the scientific breakthroughs themselves.