Finding. Major depressive disorder and schizophrenia do not share a DSM chapter. In serum, Luo, Maes and colleagues say they share a neuroimmune-metabolic-oxidative (NIMETOX) architecture. Eighty-nine people (26 healthy, 34 depression, 29 schizophrenia). Untargeted metabolomics: 1,419 features, 38 selected markers, 11 modules.
Both patient groups, versus healthy controls, show weaker mitochondrial lipid bioenergetics and weaker metabolic resilience, and more sphingolipid stress signaling, oxidative lipid damage, and renin-angiotensin-aldosterone stress. Polyunsaturated-fatty-acid membrane remodeling climbs from healthy to schizophrenia to depression. Antioxidants fall in depression only. A nested cross-validated orthogonal partial least-squares model separates healthy people from pooled patients with area under the curve 0.946. Across dimensional regressions, the mitochondrial lipid, polyunsaturated-fatty-acid, sphingolipid, oxidative-lipid, and resilience modules explain substantial variance in mood, positive and negative psychotic symptoms, resilience, and suicidal behavior. A terpenoid exposure / biotransformation module associates strongly with psychosis and suicidality.
Why this paper matters
The mitochondrial claim is specific enough to use: a lipid-bioenergetics module that moves the same way in two diagnoses and then tracks dimensions. That is more interesting than "mitochondria are involved in psychiatry." It also refuses the diagnostic binary. If modules explain mood and suicidality across labels, the blood signature is closer to a dimension than to a disorder.
The 0.946 figure will be quoted. Quote it correctly. It is healthy versus pooled patients, not a blood test that tells depression from schizophrenia.
What they actually measured
Serum metabolomics, modular collapse, group tests, a discriminant model, and multiple regressions onto clinical dimensions, childhood adversity, and resilience. The abstract does not list the 38 markers, does not define mitochondrial lipid bioenergetics chemically, and does not report medication or body-mass control.
Treat the mitochondrial module as a serum lipid pattern with that name, not as a biopsy of neuronal oxidative phosphorylation. n=89 is small for 1,419 features; the authors already collapsed to 38 markers, which is where overfitting usually hides.
How to read the score
Low seventies. Named mitochondrial module, transdiagnostic direction, dimensional regressions, a quoted discrimination statistic. Confidence is medium because the design is low-N, serum-only, and marker-selected. It is not a mechanistic mitochondrial paper.
What to do with it
If you build psychiatric metabolomic classifiers, pull the 11-module definitions and the dimensional regressions, especially mitochondrial lipid bioenergetics versus suicidality. Replicate before you ship a panel. Do not diagnose from this brief.