Bravizine: A β-Arrestin–Biased Partial Agonist Targeting Dopamine D2 Receptor Hypersensitivity in Tardive Dyskinesia

Tardive dyskinesia (TD) arises from chronic dopamine D2 receptor (D2R) blockade, leading to receptor hypersensitization and maladaptive downstream signaling. To address this, we designed Bravizine, a novel -arrestin–biased partial agonist, using an AI-augmented ligand optimization workflow built on REINVENT 4 with custom scoring functions for receptor bias prediction, docking affinity, and drug-likeness. Molecular docking predicted high-affinity binding of Bravizine to D2R (−10.8 kcal/mol, 6CM4 template), with conserved interactions at Asp114 and Phe389 and minimal predicted off-target affinities at opioid and serotonergic receptors (>−7.5 kcal/mol). Molecular dynamics simulations (50,000 ns) revealed a stable Bravizine–D2R complex characterized by conformational rearrangements in ICL2 and ICL3, structural hallmarks of effector bias. Arrestin2 docking (HDOCK) yielded a score of −291.9 with confidence 0.94 and maintained an average of 300 hydrogen bonds across frames, indicating stable -arrestin complex formation. Bias prediction models further supported a -arrestin–dominant signaling profile, with suppressed G-protein interaction energetics relative to quinpirole control simulations. Transcriptomic analysis of human frontal cortex datasets (GSE174407) indicated elevated DRD2 expression in antipsychotic-treated schizophrenia samples (fold change +1.6, adjusted p<0.05). In silico Bravizine modulation predicted partial normalization of DRD2 -associated pathways, consistent with a receptor desensitization mechanism. Bravizine displays a favorable physicochemical and pharmacological profile (QED=0.90, MW = 412 Da, LogP = 3.2, TPSA = 72 Å​2), high on-target affinity, and robust predicted -arrestin bias. These findings, derived entirely from computational modeling, support Bravizine as a rationally designed, first-in-class therapeutic candidate targeting D2R hypersensitization in tardive dyskinesia and justify future experimental validation.