Background: Glutamatergic dysfunction has been implicated in schizophrenia. Peripheral metabolic enzymes, including glutamate-oxaloacetate transaminase (GOT/AST) and glutamate-pyruvate transaminase (GPT/ALT), regulate systemic glutamate homeostasis and may serve as accessible biomarkers. This study examined the association between circulating liver transaminase levels and psychotic symptom severity in schizophrenia.Methods: A retrospective observational study was conducted using electronic medical records from patients with schizophrenia (N = 218), yielding 436 observations across ambulatory and hospitalization visits. Symptom severity was assessed using retrospectively assigned SCH-CGI scores and dichotomized into low versus high severity. Predictors included GPT and GOT log-transformed Z-scores, the log-transformed GOT/GPT ratio, age, and sex. Machine learning models (logistic regression, RBF-SVM, and gradient boosting) were applied using group-aware data splitting, and performance was evaluated using ROC-AUC and PR-AUC.Results: Biochemical predictors demonstrated moderate but consistent ability to predict symptom severity among paired hospitalization and ambulatory states patients with schizophrenia. Logistic regression including GPT and GOT log-Z scores achieved the highest performance (ROC-AUC = 0.74; PR-AUC = 0.79), outperforming more complex models. The GOT/GPT ratio did not add predictive value.Conclusions: GOT and GPT were associated with psychotic symptom severity, supporting a potential role for peripheral glutamate metabolism in psychosis. Although predictive performance was modest, these findings highlight the relevance of accessible metabolic markers within multimodal biomarker frameworks and warrant validation in larger prospective studies.