Multiscale h-Foci and CNS-Gated Reversal Identify Riluzole as a Repurposing Candidate in Alzheimer's Disease

Transcriptomic reversal and molecular docking are widely used in drug repurposing, yet both can remain positive when clinical efficacy is absent. We integrated three Alzheimer’s disease brain cohorts (163 cases, 103 controls), a signed continuous-h network, donor-level single-nucleus RNA sequencing, neural LINCS L1000 profiles, Drugs@FDA, a transparent CNS screen and exact human co-structure redocking. The analysis yielded 3,209 robust genes and five data-adaptive h-foci; only the excitatory-neuron signature survived global correction. From 2,321 LINCS perturbagens, 143 reversed the signature in neural contexts, 29 matched active FDA ingredients and 19 passed CNS gating. Translation review declined to promote 17 of these 19 (89.5%). Riluzole emerged by blind computational convergence and aligned with a prior phase-2 FDG-PET signal. Varenicline passed reversal, CNS and structural tests but had failed an Alzheimer’s trial, providing an explicit computational-positive–clinical-negative calibration standard. The framework separates candidate prioritization from efficacy claims.