Metabolic connectomes detect network disruption before first-episode psychosis
Whole-brain MRSI is beginning to expose a layer of brain organization that conventional imaging can miss. In this preprint, investigators applied proton magnetic resonance spectroscopic imaging across the brain to build metabolic connectomes in people at high risk for psychosis and compared those maps with diffusion MRI and resting-state functional MRI. The result was a selective metabolic signal: a significant connectivity subnetwork was stronger in the high-risk group, while structural and functional connectivity did not differ in parallel. For a condition defined by subtle early change, that separation is clinically meaningful.
The study grouped 29 individuals meeting DSM-5 criteria for Attenuated Psychosis Syndrome or Schizotypal Personality Disorder into a single high-risk cohort and compared them with 30 healthy controls. By combining three imaging layers, the team asked whether metabolic relationships between brain regions carry information that conventional connectomic approaches may overlook. They did not find corresponding abnormalities in regional metabolite concentrations, which means the signal was not a simple matter of one compound being globally elevated or depleted in a specific area. Instead, the alteration emerged at the network level, where brain regions appear to be coupled differently even when their individual metabolic values remain within the expected range.
This pattern fits the biology of psychosis risk more closely than a single-region defect would. Early psychiatric syndromes often reflect disturbed coordination across distributed circuits rather than focal injury, and metabolic coupling may capture that distributed vulnerability. A normal-appearing regional concentration can still sit inside an abnormal network architecture. In practical terms, the brain may be miscommunicating before it looks overtly damaged.
A selective subnetwork, not a broad structural collapse
Network-based statistics identified one metabolic connectivity subnetwork with higher connectivity in the high-risk group than in healthy controls. The same comparison did not produce significant differences in structural or functional connectivity. This divergence matters because it suggests that metabolic organization may be picking up a different dimension of disease biology than tract-based or resting-state measures.
The nodes inside the altered metabolic subnetwork showed modestly higher structural strength, but they were not consistently enriched among structural hubs. In other words, the affected nodes were not simply the most heavily connected anatomical relays in the brain. Their importance came from how metabolic relationships clustered across the network, not from a gross change in structural centrality. That pattern argues for a systems-level disruption that is more subtle than a lesion and more specific than diffuse instability.
One sentence captures the core finding: the abnormality appeared in the metabolic wiring diagram, not in the structural or functional maps.
Gradient coherence points to altered large-scale cortical development
The most conceptually important result came from eigenmode analysis. Metabolic connectivity followed a low-dimensional organization spanning a posterior unimodal-to-transmodal axis and a transmodal-to-limbic axis. Those axes resemble known large-scale cortical gradients, which organize brain systems from sensory-specialized regions toward association cortex and limbic territories. In healthy neurodevelopment, these gradients help impose an orderly hierarchy on cortical communication.
Path analysis then showed higher bilateral energy in the high-risk group, a pattern interpreted as reduced coherence along a metabolic gradient that parallels the neurodevelopmental sensorimotor-to-association axis. A metabolic gradient with lower coherence suggests that the brain’s large-scale organization may be less efficiently aligned than in controls. The finding does not claim psychosis is reducible to one gradient, but it does place risk biology inside a broader developmental framework: when gradient architecture is less coherent, distributed circuits may become less synchronized in the ways that support stable cognition and perception.
This is where the imaging result becomes more than a statistical contrast. Psychosis risk has long been linked to aberrant maturation of cortical association systems, and a gradient-based metabolic phenotype offers one plausible mechanistic bridge. Metabolite connectivity may reflect how energy handling, synaptic function, and network integration co-vary across the cortex during development. If that co-variation becomes less orderly, the downstream effect could be a brain that is still structurally intact but biologically less coordinated.
Why the absence of regional metabolite changes matters
The lack of significant regional abnormalities in individual metabolite concentrations is not a dead end; it is part of the story. It implies that whole-brain MRSI may be more informative when interpreted as a connectomic tool rather than as a catalog of isolated metabolite elevations or deficits. A single voxel or region may look unremarkable even while the relationships among many regions shift in a reproducible way. For psychosis risk, that distinction is especially relevant because the prodromal phase is often characterized by network fragility rather than overt tissue loss.
The study also shows why combining modalities can be useful even when only one modality yields a positive result. Diffusion MRI and resting-state functional MRI did not reveal corresponding abnormalities in this cohort, yet the metabolic connectome did. That does not make the other modalities less valuable; it suggests they answer different biological questions. Structural imaging describes wiring, functional imaging describes activity correlations, and spectroscopy-based connectomics may capture a biochemical coordination layer that sits between them.
What this preprint can and cannot establish
The cohort was relatively small, with 29 high-risk participants and 30 controls, so the work should be read as an early systems-neuroscience signal rather than a clinical biomarker ready for deployment. The high-risk group also pooled two DSM-5 categories, Attenuated Psychosis Syndrome and Schizotypal Personality Disorder, which may share vulnerability but are not identical clinical entities. Because the article is a preprint, the findings still await peer review and external replication.
Even so, the message is unusually coherent. Whole-brain metabolic connectomics detected network disruption where structural and functional analyses did not, and the alteration aligned with gradient-based organization that is biologically plausible for early psychosis vulnerability. If future studies reproduce this pattern in larger and more clinically diverse samples, MRSI-based connectomics could become a candidate method for probing the earliest network shifts in psychosis risk. For now, the result is narrower but concrete: altered metabolic network coherence may be one of the earliest measurable signatures of a brain moving toward psychosis, even when conventional MRI still looks quiet.