8. THE METABOLIC BRAIN INDEX: A HERITABLE SIGNATURE OF BIOENERGETIC VULNERABILITY

Background The brain’s energy-demanding association circuits operate at higher metabolic costs than sensory regions, with little spare capacity to absorb additional bioenergetic demand. Linking this bioenergetic vulnerability to brain disorders has so far required FDG-PET, which does not scale to large samples. Brain MRI does scale but is metabolically uninformative. We introduce the Metabolic Brain Index (MBI), which aggregates MRI-derived regional normative deviations, weighted by a canonical FDG-PET map of local cerebral glucose metabolism. It is constructed to be phenotypically orthogonal to an unweighted Global Brain Index (GBI), which captures overall brain deviation. As such, the MBI captures the spatial signature of brain deviation that aligns with regional metabolic demand. Methods We computed MBI in 83,271 UK Biobank brain imaging participants using a 70/30 discovery/replication split, with normative models trained on the discovery sample. To establish its clinical relevance and biological substrate, we tested the MBI's associations with depression phenotypes, as well as cardiometabolic and cognitive domains; ran null-weighting permutations to assess specificity to the FDG hierarchy; and performed GWAS in 69,024 unrelated individuals, with downstream genetic-architecture, pathway, and cell-type analyses. Results MBI tracked depression risk in a severity-graded manner, with hospitalised depression showing the strongest effect (OR=1.15 per SD, P=5.7 × 10⁻¹³), antidepressant use intermediate (OR=1.10, P=8.0 × 10⁻¹²) and broad self-reported symptoms the weakest (OR=1.03, P=0.048); the two strongest signals replicated in the held-out sample. The canonical FDG weighting outperformed all 1,000 random permutations of the PET map for hospitalised depression (empirical P=0.001), establishing that the FDG hierarchy specifically carries this signal; negative control phenotypes were null. Cardiometabolic and cognitive associations were broad and replicated in the hold-out sample, with metabolic effects systematically stronger in younger participants and cognitive effects stronger in older. The MBI GWAS identified 72 genome-wide significant loci with SNP-based heritability h²=0.303 (SE 0.015) and was uncorrelated with GBI (rg=.03), confirming a genetically distinct dimension. Variant-to-gene prioritisation across the lead variants converged on capillary morphogenesis and cytoskeletal organisation, most prominently Hedgehog signalling. Cell-type stratified MAGMA localised the genetic signal to brain microvascular endothelial cells (combined P=1.3 × 10⁻⁹, FDR=1.8 × 10⁻⁸), consistently across cortex, hippocampus, thalamus and cerebellum. Discussion MBI's genetic architecture converges on cerebral capillary biology, the rate-limiting cellular layer through which glucose enters brain tissue via endothelial GLUT1. By isolating bioenergetic vulnerability from general brain deviation, this approach provides a scalable route from population MRI to the metabolic basis of psychiatric risk.