AI‑Driven Integrative Design of ADAMTS Partial Agonists within a Multi‑Hallmark Therapeutic Framework for Neurodegenerative Disorders

Neurodegenerative diseases (NDDs) are marked by progressive neuronal dysfunction, synaptic loss, and imbalances in extracellular matrix (ECM) remodeling driven by aberrant ADAMTS‑4/5 activity. Here, we present and suggest an AI‑driven retrosynthesis framework that leverages large language models (LLMs), AlphaFold2‑predicted enzyme structures, and MolGPT generative chemistry to rediscover and optimize partial agonists of ADAMTS enzymes. Our workflow suggest the using LLMs to mine structure–activity relationships, to propose scaffold and bioisosteric modifications, and automated docking into predicted ADAMTS active sites to triage candidates with balanced binding affinity and functional modulation. Compounds are evaluated against key in silico metrics—binding energy, bioisostere similarity, solvation profile, and docking potential—to retain physiological ECM turnover while damping pathological proteolysis. This integrative approach enables selective modulation of ADAMTS activity, restoring ECM integrity and attenuating neuroinflammation, thereby offering a scalable path toward functional cures for Alzheimer’s, Parkinson’s, and related disorders ​