Ordered gene similarity emerges along the human DCML pathway
A human transcriptomic screen has identified a fixed 19-gene panel whose regional expression pattern tracks the anatomical sequence of the dorsal column-medial lemniscus system. The work asked a narrow but biologically important question: do molecular signatures in the brain preserve the same ordered relationships seen in long-range sensory pathways? In the somatosensory system, the answer was yes often enough to stand out against multiple null models, even if the strongest statistical case depended on how the comparison was framed.
The study began with neuroanatomical pathways drawn from the literature and matched them to available tissue representations in the Body Index of Gene Expression, a human microarray resource spanning cells and tissues. Investigators screened 10,262 genes under two practical constraints: a candidate had to appear across every mapped region of a route, and it could not be dominated by a single region. Those rules quickly narrowed the field. Twenty-three candidates survived in three routes with sufficient coverage, and 19 of them satisfied the criteria for the DCML representation, forming the fixed panel used for downstream testing.
One sentence captures the biology behind the screen: the genes were not selected for simply being present, but for preserving anatomical order across connected relay stations.
A pathway test centered on adjacent regional relationships
The panel was then examined in six Allen Human Brain Atlas donors across three central tissue representations: an aggregate of samples assigned to the cuneate and gracile nuclei, the lateral group of nuclei in the ventral division of the dorsal thalamus, and the postcentral gyrus. The key metric was a regional-order contrast. It compared the mean similarity of the two adjacent pairs in the anatomical sequence — cuneate/gracile nuclei to dorsal thalamus, and dorsal thalamus to postcentral gyrus — against the similarity between the two endpoints, cuneate/gracile nuclei and postcentral gyrus.
Both adjacent-in-order correlations were positive in all six donors. The contrast was positive in five of the six and produced a mean T of 0.345175 with an exact sign test P value of 0.109375. In plain clinical terms, the molecular profile became more similar along the expected anatomical route than it did when the first and last stations were compared directly. In a sensory relay system built to preserve and sharpen afferent information, that pattern is biologically plausible. Peripheral input travels through the dorsal columns, synapses in the medulla, ascends through the medial lemniscus, and reaches the thalamus before projecting to cortex. A transcriptomic gradient aligned with that order suggests that tissue identity may still carry a footprint of pathway architecture.
Null models moved the signal between suggestive and statistically stronger
The most striking part of the analysis was not the direction of the effect, but how much the inference shifted with the comparator. The 19-gene panel exceeded two distributions of 50,000 gene sets generated under comparable selection constraints, with P values of 0.027919 and 0.026579. In descriptive anatomical comparisons, it ranked third among 240 combinations and first among the target and all 16 single-region substitutions. Every one of 50,000 one-probe-per-gene analyses retained T above zero, and broad neuronal-expression contrasts were positive in medulla, thalamus, and cortex.
The panel’s support also varied by null model. Under whole-donor surrogates, the spatial signal was stronger, with P = 0.003860. Under the primary within-compartment model, the evidence weakened to P = 0.051179. This split matters because it tells readers where the apparent order is robust and where it sits near the threshold of conventional significance. The biology is still coherent, but the statistics are not equally unforgiving across all testing frameworks.
Why ordered transcriptomic similarity matters in somatosensory anatomy
The central concept here is modular continuity. Neural pathways are not just wiring diagrams; they are collections of relay nuclei and projection zones that may retain shared transcriptional programs as they pass information across the neuraxis. In the DCML system, fine touch and proprioceptive signals move through tightly organized stations that differ in function, cellular composition, and synaptic architecture. A gene panel that follows that progression could reflect shared developmental patterning, conserved neuronal populations, or molecular features tied to axonal connectivity and relay-specific physiology.
The current dataset does not prove any single mechanism. It does, however, define a specific hypothesis: anatomically ordered somatosensory relays may be recoverable from a small, fixed transcriptional signature rather than from broad, unsupervised gene lists. That has implications for how future atlases are built. If the finding holds in independent, better-resolved tissue maps, transcriptomic ordering could become a tool for validating pathway anatomy, comparing homologous nuclei, or flagging mislabeled samples in human brain reference datasets.
The study also carries a caution familiar to anyone working at the interface of anatomy and omics. A signal that emerges in one donor set can still depend on how the null is built, how regional samples are aggregated, and how strongly the analysis penalizes local similarity versus pathway order. Here, the evidence was consistent in direction across donors, but the strength of support varied enough to keep the conclusion appropriately bounded.
What this preprint can and cannot claim today
This is a preprint from bioRxiv, not a peer-reviewed clinical study, and its value lies in hypothesis generation rather than practice change. Even so, it is unusually concrete for a transcriptomic analysis because it does not stop at a diffuse enrichment result. It identifies a fixed 19-gene panel, ties it to a named sensory pathway, and shows how the signal behaves under multiple inferential settings.
Future confirmation will need an independent anatomically resolved dataset with matching regional granularity, particularly for the medullary, thalamic, and cortical nodes used here. It will also need direct assessment of whether the panel reflects neuronal subtype composition, developmental patterning, or projection-related molecular architecture. Without that step, the result remains a precise but still provisional map of how human somatosensory tissue may encode pathway order at the transcriptomic level.