Donor-level robustness and limited cross-cohort transportability of COX15-positive fibro-adipogenic progenitor states in Duchenne muscular dystrophy

AbstractBackground: Fibro-adipogenic progenitors (FAPs) are heterogeneous stromal cells thatsupport repair but can also contribute to fibrofatty replacement in Duchenne muscular dystrophy(DMD). A rare FAP state with detectable COX15 RNA was nominated in earlier exploratorymapping. We tested whether its disease-associated direction was robust to FAP definitions,transported across donors and cohorts, and could support a practical spatial-validation design.Results: We reanalyzed public raw single-nucleus RNA-sequencing counts using a frozendonor-level protocol without imputation. In the primary cohort, 61,529 quality-controlled nucleifrom 7 DMD and 5 control donors were retained. Under the primary competitive FAP gate withexclusion of the highest-remodeling quartile, 136 of 2,966 FAPs had detectable COX15. Meandonor fractions were 4.63% in DMD and 2.89% in controls (difference 1.75 percentage points;ratio 1.50; exact donor-label permutation P=0.134), with a positive leave-one-donor-out directionin 12/12 analyses. The DMD-control direction was positive in all 12 prespecified computablegate-by-threshold combinations, although only 3/12 yielded nominal P<0.05. In a contextualcohort of 61,089 nuclei from 3 DMD, 3 Becker muscular dystrophy, and 5 control donors, theDMD-control effect was non-significant and gate dependent. CCDC25 was associated withCOX15 detection in discovery (odds ratio 5.73; false-discovery-rate q=0.0066) but failed nesteddonor prediction and reversed direction externally. A two-section four-plex RNAscope designanchored by DCN was therefore prioritized. Simulations indicated approximately 20 donors pergroup for 80% power when 200 analyzable FAPs are counted per donor.Conclusions: COX15 detection marks a rare, context-dependent FAP transcriptional staterather than a validated diagnostic biomarker. Its direction is robust within the primary cohort, butits amplitude is not yet transportable across cohorts. Independent tissue and donor-level spatialcounting are required before functional or clinical interpretation.