A multimodal analysis suggests partial IRF8/TLR7-associated myeloid transcriptional convergence between diabetic kidney disease and Parkinson's disease.

BACKGROUND: Diabetic kidney disease (DKD) and Parkinson's disease (PD) affect different organs but share epidemiological associations and innate immune abnormalities. The extent and cellular basis of reproducible cross-disease transcriptomic convergence remain uncertain. METHODS: Two DKD glomerular and three PD substantia nigra microarray cohorts were analyzed independently and combined using disease-specific and cross-disease meta-analysis. Independent DKD single-cell RNA-sequencing, PD single-nucleus RNA-sequencing, and spatial transcriptomic datasets were used for cell-type localization and donor- or sample-level analyses. CellChat, Monocle2, scTenifoldKnk, and previously generated UNAGI-compatible outputs were used to evaluate predicted intercellular communication, transcriptional-state continua, in-silico IRF8 perturbation, and compound prioritization. Threshold, leave-one-cohort-out, root-orientation, downsampling, and statistical-unit sensitivity analyses were performed where applicable. RESULTS: The original nominal-threshold intersection of 36 genes was not retained after cohort-wise multiple-testing correction. Genome-wide overlap between DKD and PD was limited in the expanded analysis. IRF8 and TLR7 showed positive effect directions in both diseases but did not pass cross-disease false-discovery-rate correction and were therefore treated as exploratory candidates. Donor-level analyses did not detect statistically significant increases in broad myeloid-lineage proportions, although the confidence intervals did not establish equivalence. Selected DKD CCL and TNF interactions were retained in both of two targeted balanced-downsampling iterations, whereas the PD TGF-β signal was sampling-sensitive. Monocle2 identified branched and partially overlapping transcriptional-state continua. Spatial analyses provided marker-supported localization evidence without treating individual spatial observations as independent replicates. In-silico IRF8 deletion predicted regulatory-network changes but was not experimentally validated. NVP-AUY922 was the highest-ranked cross-disease computational candidate, but its association was not significant after correction across the matched compound universe. CONCLUSIONS: DKD and PD show limited genome-wide overlap but partial convergence of myeloid- and microglia-enriched inflammatory signals associated with IRF8 and TLR7. These findings provide testable hypotheses rather than evidence of a conserved causal pathway or validated cross-disease treatment.