Blind cross-disease topological analysis: Alzheimer's vs Parkinson's disease single-cell RNA-seq — DON Research engine v1

Blind topological collapse analysis comparing Alzheimer's disease (AD) and Parkinson's disease (PD) across three published human single-cell / single-nucleus RNA-seq datasets, performed by the DON Research engine on 2026-03-15. Datasets analyzed:- GSE138852 (Grubman et al., Nature Neuroscience 2019): 13,214 cells, human brain, AD vs control- GSE157783 (Smajić et al., Brain 2022): 41,434 cells, human midbrain, PD vs control- GSE202210: ~78,000 cells, human prefrontal cortex, PD vs control Methodology: cell type labels were taken from the dataset authors' published metadata (standard scRNA-seq cell typing). Disease labels (AD/PD/control) were NOT provided to the engine during collapse — they were used only at the comparison stage to compute Phi energy ratios per cell type. The DON-GPU collapse produces a Phi energy value, coherence, spectral entropy, and basin coordinates per cell-type aggregate; ratios are PD-or-AD divided by matched control. Headline finding — opposite microglial energy states across diseases:- AD microglia: Phi ratio 0.88× control (depleted)- PD midbrain microglia: Phi ratio 1.07× control (activated)Same cell type, opposite topological direction across two independent diseases. Microglial sample sizes: PD n=2698 cells, control n=1205 cells (GSE157783); AD: full Grubman cohort. Additional findings (with sample sizes and limitations disclosed in the source files):- PD dopaminergic neurons hyperactivated (Phi ratio 1.09×, n=42 PD / 32 control — small sample, treat as suggestive)- AD oligodendrocytes most-affected cell type (0.80×); PD oligodendrocytes moderately affected (0.94×)- AD-specific OPC compensation (1.04×); not present in PD (0.98×)- ALDH1A1 (dopaminergic identity gene) loss across both PD brain regions (LFC -3.77 midbrain DaNs, -1.19 cortex bulk)- PD-specific CADPS2+ glutamatergic interneuron population (117 PD vs 3 control cells)- Cortical immediate early gene storm in PD (NPAS4, FOS, FOSB, JUNB, ARC all elevated; chaperone system depleted) Files:- pd_midbrain_collapse_2026-03-15_203417.txt — GSE157783 whole-dataset and per-cell-type collapse (PD vs control)- pd_deep_dive_2026-03-15_204112.txt — gene-level analysis of DaNs, microglia, excitatory neurons, oligodendrocytes, CADPS2+ neurons, per-patient Phi- pd_cortex_collapse_2026-03-15.txt — GSE202210 cortex collapse with phase-transition markers and spectral comparison- cross_disease_analysis_2026-03-15.txt — cross-disease summary with energy landscape table and biological interpretation This deposit externalizes the analysis artifacts produced on 2026-03-15 to provide a third-party-verifiable timestamp via Zenodo's CERN-backed archival infrastructure. Files are byte-identical to original generation. Limitations: dopaminergic neuron sample sizes (n=42/32) are small and the 1.09× ratio should be treated as suggestive rather than definitive. Microglial findings carry stronger statistical weight given larger sample sizes. Biological interpretations of gene signatures are presented as connections to published literature, not as formal verification — a separate verification artifact with structured citations is in preparation.