Experts highlight biomarker-driven schizophrenia care—computational subtypes, muscarinic drugs, sharper negative-symptom scoring, and KCC2 targets—reducing trial-and-error prescribing.
How can clinicians move schizophrenia treatment from trial and error toward a
principled, mechanism-based match between patient and medication? At the inaugural Precision Therapeutics in Schizophrenia symposium, Michael Halassa, MD; Steve Brannan, MD; Srihari Gopal, MD, MHS; and Zhong Zhong, PhD; sought to answer this question. Halassa framed the problem frankly: an estimated 30% of patients with schizophrenia do not have medications that adequately address their symptoms.1
Halassa argued that psychiatry lacks the objective biomarkers available in fields like cardiology, forcing clinicians to combine noisy clinical measurements into idiosyncratic treatment decisions. He reframed precision psychiatry as an engineering problem: reducing a high-dimensional feature space into a smaller set of latent variables that map onto specific treatment targets. Those latents, he said, should cluster around 3 disease-generative processes relevant to psychotic disorders—model building, model revision, and credit assignment—which correspond loosely to frontal-cortical and striatal circuits. "If you take seriously the idea that the brain runs algorithms just like machines do, then you have the opportunity to put people in structured tasks and environments to expose that algorithm," Halassa said.¹ Using a cue-switching task administered to patients with schizophrenia and healthy controls, his lab identified 2 patient subtypes based on how sensory uncertainty was incorporated into decision-making; the subtypes did not differ in positive symptom burden but did differ in negative symptoms.
Brannan then detailed the development of xanomeline-trospium (Cobenfy), approved for schizophrenia in 2024 as the first new antipsychotic mechanism in roughly 70 years. Xanomeline, derived from arecoline found in betel nuts, produced dose-dependent psychosis remission in a 1990s Alzheimer disease trial but was shelved due to cholinergic adverse events. Pairing xanomeline, which crosses into the brain, with trospium, which does not, cut adverse events roughly in half while preserving central efficacy at the M1 and M4 muscarinic receptors—M4 enriched in the striatum, and M1 in the frontal cortex, the seat of executive function. Unlike haloperidol or clozapine, which act on dopamine D2 receptors, xanomeline-trospium avoided weight gain, sedation, and extrapyramidal symptoms, Brannan emphasized. Phase 1 dosing settled on 20 mg and 40 mg of xanomeline, and phase 2 and 3 trials enrolled patients with schizophrenia in inpatient settings to ensure medication adherence. In the phase 3 EMERGENT trials, xanomeline-trospium produced statistically significant reductions in PANSS total score versus placebo, with effect sizes of 0.6 or greater.2 Post hoc analysis also showed a numerically larger placebo-controlled response among patients with prominent negative symptoms at baseline, and cognitive improvement was concentrated in the subgroup with baseline impairment.
Gopal continued with clinical insights since approval: "We would hear these reports of people who were awakening, who were brightening up, who were opening up that they hadn't been before," he said, describing patients who had been treatment-resistant for decades.2 Because the 30-item PANSS was developed in the 1980s and may not fully capture this phenomenon, a colleague, Ken Kramer, constructed a 6-item PANSS Clarity Composite combining anhedonia, social interaction, and motivation items. Across the pooled EMERGENT dataset, the composite showed a numerically larger treatment effect than PANSS total score or established factor scores such as the Marder negative factor, though Gopal cautioned the analysis was exploratory and unpublished. He noted the composite's basis was clinical intuition rather than formal factor analysis, and that confirmatory work remains needed to determine whether it captures a construct distinct from existing negative-symptom measures.
Zhong, of Ovid Therapeutics, also presented preclinical data, focusing on OV4071, an oral activator of the potassium-chloride cotransporter KCC2. KCC2 maintains the chloride gradient underlying GABAergic inhibition, and its dysregulation has been implicated in schizophrenia's excitatory-inhibitory imbalance, he explained.3 In animal models, OV4071 reduced MK-801-induced striatal dopamine release—unlike most antipsychotics, including clozapine—and normalized amphetamine-induced hyperlocomotion without the sedation produced by a comparator antipsychotic. OV4071 also reversed phencyclidine-induced social interaction deficits, comparably to clozapine, and produced EEG changes (reduced gamma-band and increased delta-band power) that Zhong's team hopes will translate as a biomarker in ongoing phase 1 human studies. Because OV4071 has only reached phase 1, Zhong emphasized that no human efficacy data yet exist, though early safety findings have been favorable.
Taken together, the 4 presenters illustrated converging strategies—computational phenotyping, muscarinic agonism, refined negative-symptom measurement, and novel circuit targets—aimed at the same goal: matching schizophrenia treatment to underlying biology rather than a “try it and see” approach.
Dr Halassa is a professor at the Fralin Biomedical Research Institute at Virginia Tech, with appointments in psychiatry and behavioral medicine at the Virginia Tech Carilion (VTC) School of Medicine and in biomedical engineering at the College of Engineering.
Dr Brannanis a psychiatrist and former chief medical officer at Karuna Therapeutics.
Dr Gopal is vice president, head of clinical development in neuropsychiatry at Bristol Myers Squibb.
Dr Zhong is chief scientific officer at Ovid Therapeutics.
References
1. Halassa M, Brannan S, Gopal S, et al. Theme I: Innovative Pharmacology and Next Generation Therapeutics. Presented at: Precision Therapeutics in Schizophrenia Symposium; September 23-24, 2026; Roanoke, VA.
2. Kaul I, Sawchak S, Walling DP, et al.
3. Arion D, Lewis DA.
