Single-Cell Transcriptomic Coupling to Whole-Brain Connectome Topology Reveals the Molecular and Evolutionary Genetic Code of Selective Hub Vulnerability

Correction notice (30 October 2026). A post-publication audit by the author found that several headline claims of this record are not supported. (1) The "Molecular Proteostasis Gap" contains connectome wiring cost and PageRank in its own formula (55% weight), so its reported correlations with wiring cost and synaptic load (ρ = 0.64) are circular; the transcriptomic component alone is not associated with wiring cost (ρ = 0.17, p = 0.13). The statement that results were "confirmed via zero-circularity partial correlations" is therefore incorrect. (2) The channel-density correlation (ρ = 0.40) was with soma-to-arbor distance, not wiring cost (ρ = −0.11). The Na+/K+-pump correlation (ρ = 0.56) is with synapse count; wiring cost adds no information beyond synapse count. (3) All in-silico "gene knockdowns" (Pink1/parkin, Appl/tau, TBPH/caz), the "Day 70 hyperexcitability" and all "rescues" (including the 97.7% Atg1/Atg8a mitophagy rescue) in the LIF simulations are direct consequences of hand-set parameters (which neurons and what fraction are removed, synaptic scaling factors) and contain no gene-specific biology; they are withdrawn and must not be read as therapeutic predictions. The listed "neuroprotective targets" are genes more expressed in neurons with many synapses and are not supported as therapeutic targets. (4) The in vivo validation (ρ = −0.33, p = 0.048, n = 37) is marginal and does not survive correction for multiple testing. (5) A pre-registered follow-up in human data (SEA-AD, 7 cortical regions) found that human orthologs of these gene modules do not predict neuronal loss in Alzheimer's disease. (6) The previously listed co-creator "Computational Connectomics and Multi-Omic Systems Neuroscience Consortium" does not exist as an organisation and has been removed; this work has a single author. The attached PDF still contains the original, unsupported claims. A corrected version is in preparation. Audit code and results are available from the author. Corrected abstract. We matched 76 neuronal cell types between the FlyWire connectome and Drosophila single-cell RNA-seq atlases to ask whether transcriptional programs scale with connectivity. Cell types with more synapses allocate a larger share of their transcriptome to Na+/K+-pump genes (ρ = 0.56; ρ = 0.51 after controlling for transcriptome size and detected genes). Synapse-weighted wiring cost adds no information beyond synapse count. An earlier composite "proteostasis gap" index is not reported here because it contained connectome variables in its own definition. Across ageing (Aging Fly Cell Atlas, days 5–70), neuronal OXPHOS and autophagy module expression decline while core synaptic-release transcripts are maintained, consistent with previous reports. An association between an OXPHOS/autophagy ratio and in vivo transcript loss at day 50 was marginal (ρ = −0.33, p = 0.048, n = 37) and does not survive correction for multiple testing. Earlier leaky integrate-and-fire simulations of gene knockdowns and rescues reflected hand-set parameters and are withdrawn. In a pre-registered follow-up, human orthologs of these modules did not predict neuronal loss in Alzheimer's disease across seven cortical regions of the SEA-AD atlas.