Receptor-Prior Sensitivity Reveals Transancestral Stability and Clinical Discordance in HDL-Related Antipsychotic Metabolic-Risk Ranking: Insights from an Integrated Ki–DDD–TWAS Pipeline

Antipsychotic-associated metabolic abnormalities vary substantially between drugs, yet computational rankings of metabolic liability may depend on receptor priors that have not been systematically evaluated across ancestries. This study examined the sensitivity of an integrated pharmacological–genomic prioritisation framework to the use of literature-derived receptor weights. The upstream pipeline combined receptor-binding affinity, defined daily dose, and precomputed transcriptome-wide association study results for high-density lipoprotein cholesterol across five ancestry-specific datasets from the 2021 Global Lipids Genetics Consortium resource. The present Stage 3 analysis used saved receptor-level and drug-level outputs to compare a literature-weighted model with a uniform model in which all recognised receptor genes received a weight of 1.0. The analysis included receptor-class contribution shifts, leave-one-gene-out rank sensitivity, drug-specific contribution fingerprints, top-10 set stability, transancestral gene-reallocation consistency, delta-rank concordance, secondary-boost decomposition, candidate-gene screening, and comparison with a fixed clinical ordering.The literature-weighted model produced a completely stable top-10 set across the five ancestries, with a mean Jaccard index of 1.000, and showed better concordance with the prespecified clinical comparator than the uniform model. Removing the literature priors increased the pooled dopaminergic contribution from 17.70% to 44.15%, while histaminergic contribution decreased from 26.01% to 8.85%. Under weighted scoring, HRH1 was the dominant single-gene driver, with a mean absolute leave-one-gene-out rank shift of 5.01. Under uniform scoring, DRD2 became the dominant driver, with a mean absolute rank shift of 4.32. Ziprasidone rose consistently in all five ancestries, whereas levomepromazine, promazine, acepromazine, olanzapine, and quetiapine generally moved downward. All 39 genes that changed contribution did so in the same direction across ancestries, and delta-rank profiles showed a mean Spearman correlation of 0.952. The strongest high-TWAS/low-weight candidates, ADRB1 and CHRM4, were limited to the European dataset under the predefined thresholds.These findings support dual reporting: a literature-weighted model for clinically anchored prioritisation and a uniform model for sensitivity analysis and hypothesis generation. Neither model should be interpreted as a validated clinical prediction model or as evidence of a causal receptor hierarchy.