Human Neuropathic Pain and Peripheral Nerve Injury Form Largely Separable Transcriptomic Axes Disentangling symptom-associated transcription from lesion injury and remodeling across human pain datasets

Transcriptomic studies of neuropathic pain frequently analyze tissue obtained from an injured or diseased anatomical site. Thiscreates a fundamental interpretive problem: a molecular difference observed in painful tissue can encode the subjective painphenotype, the structural nerve lesion, demyelination, immune infiltration, repair and remodeling, or a mixture of theseprocesses. Here we tested whether human neuropathic-pain-associated transcription can be separated from the transcriptionalfootprint of peripheral nerve injury. The analysis arose from a falsification-driven cross-dataset study in which a prespecifiedfive-gene low-expression score (DP-LES5; SAMD12, MEGF10, CSMD2, TSPAN5 and ENOX1), initially motivated bydepression-pain convergence, was challenged in increasingly direct human pain datasets. The score was increased intrigeminal-neuralgia whole blood (loss difference +0.308; exact one-sided P=0.038) and in Morton neuroma relative to uninjuredperipheral nerve (+0.499; permutation P=0.0057), but reversed in chronic-versus-acute human C2 dorsal root ganglion (DRG)tissue (-0.825; two-sided exact P=0.0194) and was null in a larger pain-associated-versus-non-pain human DRG comparison (-0.037; sex-stratified two-sided permutation P=0.773). These contradictions falsified a universal pain-associated low-expressioninterpretation and motivated a transcriptome-wide comparison. Across 19,760 shared genes, the lesion-site Morton-neuromaeffect vector and the human DRG pain-phenotype vector showed only minimal rank concordance (Spearman rho=0.0537).Projection/residualization produced a pain-associated axis containing IL1B and other inflammatory signals and a distinctinjury/remodeling axis characterized by peripheral myelin/Schwann-cell losses (including PRX, MPZ, DRP2 and CLDN19)together with remodeling/immune signals such as MARCO. DP-LES5 components including ENOX1, TSPAN5 and MEGF10localized preferentially toward the injury/remodeling dimension rather than the direct DRG pain dimension. Our central finding istherefore that human neuropathic pain and peripheral nerve injury can occupy largely separable transcriptomic axes.Consequently, molecular changes measured in painful injured tissue should not automatically be interpreted as pain-specificmolecular signatures: a substantial component may instead encode nerve injury, demyelination, cellular-composition changeand tissue remodeling. This cross-dataset decomposition does not establish causal or universal biological orthogonality, but itprovides a falsifiable framework for distinguishing symptom-associated transcription from lesion-associated transcription inhuman pain research.