TWAS-Derived Mechanistic Leads Linking NMN-Sensitive Trafficking and Bioenergetics to Eating-Disorder Genetic Risk: RHOA and Synaptic Vesicle Dynamics as Translational Entry Points

Anorexia nervosa (AN) and binge-eating behavior (BE) are clinically distinct but partially overlapping phenotypes. Their genetic architecture includes both shared liability and disorder-specific components, yet the molecular mechanisms that distinguish restrictive from binge-related eating pathology remain incompletely defined. This gap limits mechanism-based stratification and the development of targeted interventions. Here, we examined whether gene programs identified in an NMN-related transcriptional reanalysis could intersect with the genetic architecture of AN and BE.We applied an extended multi-gene-set transcriptome-wide association study pipeline to brain-tissue S-PrediXcan meta-Z scores derived from the AN and BE genome-wide association analyses reported by Termorshuizen and colleagues. Gene sets were derived from an NMN-rescue reanalysis of GSE85718 and from KEGG-mapped biological programs described in the supplied preprint. The analysis incorporated Stouffer enrichment statistics, permutation testing, Wilcoxon testing, bootstrap confidence intervals, disease-differential tests, profile-proximity measures, family-wise false discovery rate pooling, leave-one-out influence analysis, pathway-overlap analysis, robustness scoring, and leading-edge gene prioritization.The strongest result was positive enrichment of the synaptic vesicle cycle in BE broad, with a Stouffer Z of 4.628, permutation p of 0.0179, and Wilcoxon p of 0.00229. This pathway also differed significantly between AN and BE, with a Mann–Whitney false discovery rate of 0.01069 and paired Wilcoxon p of 0.04042. Leading genes included SLC17A7, VAMP2, DNM1, ATP6V1E2, and CACNA1B, implicating vesicle loading, docking, fusion, acidification, and recycling. AN broad showed a suggestive citrate-cycle signal, with a Stouffer Z of 3.127 and Wilcoxon p of 0.02767, although the competitive permutation p was 0.1334. RHOA emerged as the highest-impact recurrent hub, appearing across endocytosis, RAS signaling, and actin-cytoskeleton pathways. It was flagged five times, had the highest mean absolute leave-one-out influence among recurrent drivers, and showed three sign flips.The findings identify testable interfaces between NAD+-sensitive bioenergetics, endosomal trafficking, and presynaptic function. RHOA, SLC17A7, VAMP2, CACNA1B, ARPC5L, OGDH, IDH3B, DLD, and CPT2 merit functional investigation. These results remain hypothesis-generating: TWAS estimates genetically predicted expression rather than measured expression, and the analysis does not establish causal effects or therapeutic efficacy of NMN or CGR-related interventions.