In order to succeed in humans, we must start in humans

A foundation model is only as good as the data it learns from.

In cancer, patients can be matched to drugs because doctors can biopsy the tumor itself. In neuroscience, the brain is the molecular ground truth for disease — and the one tissue you cannot sample in a living patient. The field has had to rely instead on proxies: blood, imaging, and clinical data. Almost no patient has a complete dataset.

Generative AI can now fill in that gap, learning across modalities from incomplete data. But only if it’s anchored in brain tissue. Without it, the proxies show only the disease’s effects, not what’s driving it.

Over the last ten years, we built the anchor. VergeDB is the largest proprietary brain tissue collection of its kind, built over ten years through partnerships with 21 brain banks, hospitals, academic medical centers, and research consortia across four continents. Every sample is profiled across multiple molecular modalities and paired with detailed clinical annotations.

How we built it

From each sample, we can gather different data modalities:

1. Whole transcriptome RNA-sequencing to capture the molecular changes in tissue in disease and healthy states, across multiple brain regions and disease states
2. Genetics (whole genome sequencing or genotyping) to capture genetic alterations upstream of disease
3. Clinical annotations, including deep clinical phenotyping, disease subtypes, symptom fields, and neuropathology
4. Multi-omics data, including proteomics and single-nuclei sequencing to capture additional depth

VergeDB now includes clinical, biomarker, and patient data from VRG50635, Verge’s first AI-discovered asset, which completed a Phase 1b in ALS in late 2025 — one of the most thoroughly phenotyped clinical and biomarker datasets in the disease.