Schizophrenia splits into genetic profiles as psychiatry gets measurable maps

Profiles, not one diagnosis

Four preprints in this batch point the same way: psychiatry and neurology are getting maps precise enough to argue with. The papers are preprints, not peer-reviewed results, and they should be read that way. But taken together they describe a shift from one diagnosis, one marker, one gene test toward measurements that split patients into groups with different outlooks. The schizophrenia paper is the clearest example. Researchers tested neurotransmitter-specific polygenic scores, genetic risk scores built around one brain chemical system at a time, against the usual genome-wide score. m:…9309

The genome-wide score showed no significant association with clinical outcomes. The pathway scores did. In the preprint, the acetylcholine pathway profile was associated with higher overall symptom severity and more negative symptoms; the glutamate profile was associated with earlier age at onset, reduced three-year remission and poorer treatment response; and the GABA risk profile was associated with three-year remission rates. The pathway profiles tracked differences in symptom severity, age of onset and treatment outcomes over time. The implication is not a test a clinic can order tomorrow. It is that schizophrenia's heterogeneity, the reason two patients with the same diagnosis can have very different courses, may be partly sortable by biology. m:…9309

A circuit for naming feelings

A second preprint maps a symptom rather than a diagnosis. Researchers used lesion network mapping in patients with gliomas, correlating lesion connectivity with scores on the Toronto Alexithymia Scale, a measure of difficulty identifying and describing feelings. Lesions linked to alexithymia mapped to a common network centered on the bilateral caudate heads and the anterior cingulate cortex, and the network converged with independent meta-analysis coordinates of gray matter volume. Difficulty identifying feelings specifically localized to the caudate-centered network. The result gives alexithymia, often treated as a vague trait, a concrete circuit spanning executive and limbic systems. m:…8967

One blood marker is not enough

A third preprint, an ONDRI study, tests whether blood markers can tell two diseases apart. Plasma p-tau217 was elevated in both Alzheimer's disease and amyotrophic lateral sclerosis, which limits its use on its own. But the ratio of p-tau217 to neurofilament light chain separated ALS from Alzheimer's with an area under the curve of 0.92, against 0.65 for p-tau217 alone. The authors also report that inferred amyloid-beta positivity in cognitively impaired patients with cerebrovascular disease may indicate Alzheimer's or cerebral amyloid angiopathy co-pathology. The lesson is the same as in the schizophrenia paper: one number rarely settles a diagnosis, and a pair of numbers can. m:…9168

From uncertain to classified

The fourth preprint shows the same logic applied to a single gene. Researchers built a high-throughput automated patch-clamp assay for SCN1A, the gene behind many cases of epilepsy and Dravet syndrome, and combined current density and channel kinetics in a Bayesian framework to judge whether a variant causes loss of function. In an Australian cohort, the method reclassified 50 percent of 16 missense variants tested, including six variants of uncertain significance moved to likely pathogenic and two moved from likely pathogenic to pathogenic. In validation, 17 of 27 true pathogenic variants showed severe reductions in current density, and 7 of the remaining 10 showed an abnormal V0.5 difference. A variant that once came back as uncertain can now come back with an answer. m:…6873

None of these papers is a clinical tool yet. All four are preprints, the cohorts are modest, and polygenic scores and blood markers both need independent replication before they change practice. But the direction is consistent: measurable maps, built from genetics, lesions, blood and functional assays, are replacing single labels. For a clinician or a diagnostics developer, the useful question is no longer whether a patient has a condition. It is which measurable subgroup they fall into, and what that subgroup predicts. m:…9309 m:…8967 m:…9168 m:…6873