Abstract molecular signaling network with AI computational layers — Zermatt AI Platform
Platform

AI/ML-Guided Compound Optimization
for CNS STING Therapeutics

Zermatt's platform is built to optimize small-molecule STING inhibitors for CNS disease. Rather than discovering new targets from omics data, we focus on a defined and biologically validated pathway — cGAS–STING — and use AI/ML to improve the drug-like properties required for brain-penetrant therapeutics.

The platform integrates generative chemistry, affinity prediction, physics-based modeling, and active learning from experimental data to advance compounds across potency, BBB penetration, metabolic stability, safety, and synthetic tractability.

Platform Workflow

A closed-loop design–make–test cycle for CNS-penetrant STING inhibitors.

01 / GENERATE
Generate
Generative chemistry models create novel STING inhibitor candidates based on cross-generational SAR from our internal compound series.
02 / PREDICT
Predict
Machine learning models rank compounds using predicted STING engagement, functional activity, CNS drug-like properties, solubility, metabolic stability, and safety liabilities.
03 / REFINE
Refine
Physics-based modeling — docking, molecular dynamics, and free-energy estimation — prioritizes compounds with stable ligand–STING interactions.
04 / TEST
Test
Top-ranked candidates are evaluated through binding, cellular activity, ADMET, safety, and CNS exposure assays.
05 / LEARN
Learn
Each experimental result feeds back into the platform, improving model performance and enabling a closed-loop design–make–test cycle.

Platform Outputs

Deliverables that feed each design cycle and partner collaboration.

Example Focus Area

AGS and Interferon-Driven Neuroinflammation

Aicardi–Goutières syndrome is a rare pediatric interferonopathy characterized by chronic type I interferon activation and severe neurological involvement.

Zermatt's lead STING inhibitor program is designed to evaluate whether CNS-penetrant STING inhibition can suppress upstream innate immune signaling in disease-relevant models.

Partner with us on CNS STING therapeutics

Additional preclinical data are available upon request for qualified partners.