In order to succeed in humans, we must start in humans validate in humans model in humans predict in humans

Introducing CONVERGE: the AI operating system for neuroscience drug development, powered by state-of-the-art models.

CONVERGE combines foundation models and state-of-the-art systems biology workflows to answer the full set of questions needed for drug developers to design successful programs that work in humans.

Surface new drug targets grounded directly in human disease biology.

CONVERGE reasons over a decade of multimodal patient data to identify causal disease signatures. For each disease in our dataset, the platform can identify which genes, when modulated, would shift a patient’s brain state toward health. The signal is surfaced from the molecular ground truth of human tissue and genetics — not extrapolated from rodents.

Our first-generation platform surfaced 280+ novel drug targets at an 83% preclinical validation rate, and two were nominated into Eli Lilly’s pipeline. CONVERGE 2.0 goes further: it ranks targets not just by disease relevance, but matches them to the patient subgroups most likely to respond.

In partnership with See our partnerships

De-risk the path from target to clinic.

Once a target is chosen, the next decisions — what preclinical model to use, whether the biology translates to the right patient populations, which indications to pursue — can determine the program’s probability of clinical success.

CONVERGE lets program teams stress-test those decisions against thousands of real human patients. We can characterize how a target pathway behaves in patient tissue, across different subgroups, and identify which preclinical models actually recapitulate the patient biology that matters. The same analysis surfaces indication expansion opportunities — where the same signature is dysregulated in a different disease or subgroup.

Match the right drug to the right patient.

In most CNS clinical trials, cohorts are defined by clinical diagnosis — ALS, Alzheimer’s, Parkinson’s — but the molecular biology underneath any single diagnosis varies dramatically from patient to patient. Real responders get lost inside heterogeneous cohorts.

CONVERGE stratifies trial populations by molecular endotype rather than diagnosis, identifying the subgroups most likely to benefit before enrollment and the pharmacodynamic biomarkers that will move earliest in those patients — so that Phase 1b/2a readouts answer the questions that actually decide a program’s fate.

Under the hood

CONVERGE is powered by our industry-leading world model for patient disease biology.

Our foundation model reasons about each patient individually. From a routine blood draw, the model infers the molecular state of that patient’s brain — a virtual biopsy — and uses that inferred state to predict how they’ll respond to a given drug. The model is trained on multimodal CNS patient data proprietary to Verge, and can be fine-tuned with program- and trial-specific data to sharpen accuracy on individual molecules.

CONVERGE can

Identify the patient subgroups most likely to benefit before enrollment.
Discover pharmacodynamic and disease biomarkers to power Phase 1b/2a readouts.
Simulate how biomarkers will move under intervention, in advance of the trial.
Surface previously unrecognized molecular endotypes.