Functional Heritability Partitioning, Multi-Tissue Transcriptomic Colocalization, and Pharmacogenomic Target Deconvolution Across 127 Alzheimer's Disease Susceptibility Loci
- 1Universidad de Oviedo
Abstract
Background: The landmark genome-wide association meta-analysis of Alzheimer's disease (AD) and AD-by-proxy across N = 2,815,127 individuals (Uffelmann et al., Nature, 2026) expanded the known susceptibility landscape to 127 independent genomic risk loci (160 association signals) and established an autosomal common-variant SNP heritability of h2_SNP = 16.0% on the liability scale. Translating these 127 static statistical signals into pathway-stratified clinical trials requires resolving three quantitative gaps: how the non-APOE polygenic architecture partitions across distinct cellular programs, which effector genes operate via cortical parenchymal versus peripheral myeloid gene expression, and which targets possess nanomolar human bioactivity suitable for genetically stratified trials. Methods: Operating entirely on individual-genotype-free summary statistics structured along a previously established tridimensional diagnostic framework (Axis 1: Etiology and Genetic Architecture; Axis 2: Molecular Pathophysiology and Biomarker Programs; Axis 3: Neuroanatomical-Cognitive Topography; Menendez-Gonzalez, Alzheimers Dement, 2025), we ingested the complete NHGRI-EBI GWAS Catalog build GCST90695466 (8,159,496 autosomal and 160,704 X-chromosome variants). First, we performed functional heritability partitioning via stratified excess chi-square inflation and additive genetic variance (sum 2pq*beta^2) across +/-50 kb cis-regulatory windows of 50 fine-mapped effector genes organized into five mechanistic Axis 2 programs (MICROGLIAL_IMMUNE, PROTEOPATHY_AUTOPHAGY, ENDOCYTIC_VESICULAR, SYNAPTIC_CALCIUM, and MITOCHONDRIAL_OXIDATIVE). Second, we conducted Summary-data-based Mendelian Randomization (SMR) and Approximate Bayes Factor colocalization (PP.H4) across 54 human tissues in GTEx v10 (4,415 matched cis-eQTLs) alongside Open Targets v4 SuSiE fine-mapped credible sets spanning ROS/MAP dorsolateral prefrontal cortex (DLPFC), BrainSeq, primary human monocytes, and UK Biobank Pharma Proteomics Project (UKB-PPP) plasma pQTLs. Third, we queried EMBL-EBI ChEMBL v34 across 2,785 human bioactivity assays (IC50, Ki, Kd, EC50) and Open Targets tractability matrices to construct an empirical pharmacogenomic target atlas. Results: Across 8,159,496 autosomal variants, the genome-wide mean test statistic was chi2_GW = 1.5099. Outside the APOE/TOMM40 super-locus (max chi2 = 3,267.15; 2pq*beta^2 = 0.0882, accounting for 74.82% of top-locus additive variance), 89,589 cis-regulatory SNPs across 48 non-APOE effector genes exhibited a 12.88-fold heritability enrichment over the autosomal background. Normalized across the 16.0% polygenic liability scale, non-APOE functional heritability concentrated in MICROGLIAL_IMMUNE (39.22% share; h2_SNP = 6.28%; 13.02x enrichment; mean chi2 = 7.64) and PROTEOPATHY_AUTOPHAGY (36.34% share; h2_SNP = 5.81%; 14.98x enrichment; mean chi2 = 8.64), followed by ENDOCYTIC_VESICULAR (16.71% share; h2_SNP = 2.67%; 12.69x enrichment; mean chi2 = 7.47), SYNAPTIC_CALCIUM (5.40% share; h2_SNP = 0.86%; 10.05x enrichment; mean chi2 = 6.13), and MITOCHONDRIAL_OXIDATIVE (2.33% share; h2_SNP = 0.37%; 5.24x enrichment; mean chi2 = 3.67). Multi-tissue SMR and colocalization resolved a striking divergence between Axis 3 cortical parenchyma and myeloid compartments: QPCT (rs2300880; Z_SMR = -4.62, p_SMR = 3.93e-6, PP.H4_cis = 0.919, PP.H4_local = 0.994 in Brain_Cortex; PP.H4_local = 0.979 in Brain_Frontal_Cortex_BA9) and AXL (rs12979764; p_SMR = 2.13e-4, PP.H4_cis = 0.902, PP.H4_local = 0.924 in Brain_Cortex) colocalized directly within human cerebral cortex and ROS/MAP DLPFC. Conversely, canonical myeloid and endocytic drivers (BIN1, CD33, TREM2, INPP5D, SYK) exhibited near-zero bulk cortical eQTL colocalization (PP.H4_cis <= 0.025 in Brain_Cortex) despite high SuSiE Locus-to-Gene scores (L2G = 0.875-0.968) and robust colocalization in primary monocytes and spleen (INPP5D PP.H4_local = 0.992; BIN1 PP.H4_local = 0.999). ChEMBL v34 bioactivity profiling confirmed low-nanomolar human target potency for clinical-stage compounds across all five programs, including isradipine (2.0 nM) and nilvadipine (19.0 nM) for CACNA1S, omaveloxolone (2.2 nM) for KEAP1, fostamatinib/R406 (41.0 nM) for SYK, and varoglutamstat/PQ912 (61.0 nM) for QPCT. Conclusion: Over three-quarters (75.56%) of the non-APOE common-variant heritability of Alzheimer's disease resides within microglial immune surveillance and proteopathic clearance networks. If prospectively validated in genotype-stratified clinical cohorts, these empirical summary-statistics findings suggest that therapeutic efficacy in sporadic AD may depend on matching pathway-selective drugs to a patient's individual Functional Pathway Burden Score rather than administering uniform monotherapies to genetically unstratified populations.
Related
- Comprehensive Transcriptomic Analysis with Deconvolution and In Vivo Experiments Reveal the Role of the NLRP3 Inflammasome in Crescentic Glomerulonephritis.
- Serum Sphingomyelin and Osteoarthritis Risk: Evidence from Mendelian Randomization, Genetic Colocalization, and a Rat Model
- Pharmacogenomic predictor of long-term residual chemotherapy-induced peripheral neuropathy in ovarian cancer survivors: A substudy of the GINECO Vivrovaire study
- Pharmacogenomic determinants of antituberculosis drug exposure and toxicity in African populations: a systematic review protocol
- Estrous Cycle-Dependent Expression of Nucleobindin-2/Nesfatin-1 in the Porcine Hypothalamus: Colocalization with GnRH and Kisspeptin
BibTeX
@misc{menendezgonzalez2026functional,
title = {{Functional Heritability Partitioning, Multi-Tissue Transcriptomic Colocalization, and Pharmacogenomic Target Deconvolution Across 127 Alzheimer's Disease Susceptibility Loci}},
author = {M. Menendez-Gonzalez},
howpublished = {medRxiv},
year = {2026},
month = nov,
doi = {10.64898/2026.11.04.26366951},
url = {https://www.medrxiv.org/content/10.64898/2026.11.04.26366951v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Functional Heritability Partitioning, Multi-Tissue Transcriptomic Colocalization, and Pharmacogenomic Target Deconvolution Across 127 Alzheimer's Disease Susceptibility Loci
AU - M. Menendez-Gonzalez
PY - 2026
DA - 2026/11/04
PB - medRxiv
DO - 10.64898/2026.11.04.26366951
UR - https://www.medrxiv.org/content/10.64898/2026.11.04.26366951v1
AB - Background: The landmark genome-wide association meta-analysis of Alzheimer's disease (AD) and AD-by-proxy across N = 2,815,127 individuals (Uffelmann et al., Nature, 2026) expanded the known susceptibility landscape to 127 independent genomic risk loci (160 association signals) and established an autosomal common-variant SNP heritability of h2_SNP = 16.0% on the liability scale. Translating these 127 static statistical signals into pathway-stratified clinical trials requires resolving three quantitative gaps: how the non-APOE polygenic architecture partitions across distinct cellular programs, which effector genes operate via cortical parenchymal versus peripheral myeloid gene expression, and which targets possess nanomolar human bioactivity suitable for genetically stratified trials. Methods: Operating entirely on individual-genotype-free summary statistics structured along a previously established tridimensional diagnostic framework (Axis 1: Etiology and Genetic Architecture; Axis 2: Molecular Pathophysiology and Biomarker Programs; Axis 3: Neuroanatomical-Cognitive Topography; Menendez-Gonzalez, Alzheimers Dement, 2025), we ingested the complete NHGRI-EBI GWAS Catalog build GCST90695466 (8,159,496 autosomal and 160,704 X-chromosome variants). First, we performed functional heritability partitioning via stratified excess chi-square inflation and additive genetic variance (sum 2pq*beta^2) across +/-50 kb cis-regulatory windows of 50 fine-mapped effector genes organized into five mechanistic Axis 2 programs (MICROGLIAL_IMMUNE, PROTEOPATHY_AUTOPHAGY, ENDOCYTIC_VESICULAR, SYNAPTIC_CALCIUM, and MITOCHONDRIAL_OXIDATIVE). Second, we conducted Summary-data-based Mendelian Randomization (SMR) and Approximate Bayes Factor colocalization (PP.H4) across 54 human tissues in GTEx v10 (4,415 matched cis-eQTLs) alongside Open Targets v4 SuSiE fine-mapped credible sets spanning ROS/MAP dorsolateral prefrontal cortex (DLPFC), BrainSeq, primary human monocytes, and UK Biobank Pharma Proteomics Project (UKB-PPP) plasma pQTLs. Third, we queried EMBL-EBI ChEMBL v34 across 2,785 human bioactivity assays (IC50, Ki, Kd, EC50) and Open Targets tractability matrices to construct an empirical pharmacogenomic target atlas. Results: Across 8,159,496 autosomal variants, the genome-wide mean test statistic was chi2_GW = 1.5099. Outside the APOE/TOMM40 super-locus (max chi2 = 3,267.15; 2pq*beta^2 = 0.0882, accounting for 74.82% of top-locus additive variance), 89,589 cis-regulatory SNPs across 48 non-APOE effector genes exhibited a 12.88-fold heritability enrichment over the autosomal background. Normalized across the 16.0% polygenic liability scale, non-APOE functional heritability concentrated in MICROGLIAL_IMMUNE (39.22% share; h2_SNP = 6.28%; 13.02x enrichment; mean chi2 = 7.64) and PROTEOPATHY_AUTOPHAGY (36.34% share; h2_SNP = 5.81%; 14.98x enrichment; mean chi2 = 8.64), followed by ENDOCYTIC_VESICULAR (16.71% share; h2_SNP = 2.67%; 12.69x enrichment; mean chi2 = 7.47), SYNAPTIC_CALCIUM (5.40% share; h2_SNP = 0.86%; 10.05x enrichment; mean chi2 = 6.13), and MITOCHONDRIAL_OXIDATIVE (2.33% share; h2_SNP = 0.37%; 5.24x enrichment; mean chi2 = 3.67). Multi-tissue SMR and colocalization resolved a striking divergence between Axis 3 cortical parenchyma and myeloid compartments: QPCT (rs2300880; Z_SMR = -4.62, p_SMR = 3.93e-6, PP.H4_cis = 0.919, PP.H4_local = 0.994 in Brain_Cortex; PP.H4_local = 0.979 in Brain_Frontal_Cortex_BA9) and AXL (rs12979764; p_SMR = 2.13e-4, PP.H4_cis = 0.902, PP.H4_local = 0.924 in Brain_Cortex) colocalized directly within human cerebral cortex and ROS/MAP DLPFC. Conversely, canonical myeloid and endocytic drivers (BIN1, CD33, TREM2, INPP5D, SYK) exhibited near-zero bulk cortical eQTL colocalization (PP.H4_cis <= 0.025 in Brain_Cortex) despite high SuSiE Locus-to-Gene scores (L2G = 0.875-0.968) and robust colocalization in primary monocytes and spleen (INPP5D PP.H4_local = 0.992; BIN1 PP.H4_local = 0.999). ChEMBL v34 bioactivity profiling confirmed low-nanomolar human target potency for clinical-stage compounds across all five programs, including isradipine (2.0 nM) and nilvadipine (19.0 nM) for CACNA1S, omaveloxolone (2.2 nM) for KEAP1, fostamatinib/R406 (41.0 nM) for SYK, and varoglutamstat/PQ912 (61.0 nM) for QPCT. Conclusion: Over three-quarters (75.56%) of the non-APOE common-variant heritability of Alzheimer's disease resides within microglial immune surveillance and proteopathic clearance networks. If prospectively validated in genotype-stratified clinical cohorts, these empirical summary-statistics findings suggest that therapeutic efficacy in sporadic AD may depend on matching pathway-selective drugs to a patient's individual Functional Pathway Burden Score rather than administering uniform monotherapies to genetically unstratified populations.
ER -