Neurology

Neurological diseases are diagnosed late, progress unpredictably, and lack reliable biomarkers of therapeutic response. Current models can't capture what happens inside these patients over time.


We train virtual patient models on longitudinal paired cfDNA methylation, proteomics, imaging, and clinical data across many neurological disorders. The result is a foundation model that learns patient-level disease biology, outcome trajectories, and features driving disease progression — not population averages.

Our neurological model identifies multimodal features associated with disease subtypes, neurological decline, and novel biological insights  — from plasma-based biomarkers.


In Alzheimer's, our virtual patient representations trained on matched imaging and molecular data recover known pathological stages and predict progression of disease.


These are not retrospective analyses. They are trained to generalize to new patients.