Brain Tumors, AI and Psychiatry: Predicting Tumor-Associated Psychiatric Syndromes with Machine Learning and Biomarkers

Matei Șerban, Corneliu Toader, Răzvan-Adrian Covache-Busuioc

Brain tumors elicit complex neuropsychiatric disturbances that frequently occur prior to radiological detection and hinder differentiation from major psychiatric disorders. These syndromes stem from tumor-dependent metabolic reprogramming, neuroimmune activation, neurotransmitter dysregulation, and large-scale circuit disruption. Dinucleotide hypermethylation (e.g., IDH-mutant gliomas), through the accumulation of 2-hydroxyglutarate (2-HG), execute broad DNA and histone hypermethylation, hypermethylating serotonergic and glutamatergic pathways, and contributing to a treatment-resistant cognitive-affective syndrome. High-grade gliomas promote glutamate excitotoxicity via system Xc− transporter upregulation that contributes to cognitive and affective instability. Cytokine cascades induced by tumors (e.g., IL-6, TNF-α, IFN-γ) lead to the breakdown of the blood–brain barrier (BBB), which is thought to amplify neuroinflammatory processes similar to those seen in schizophrenia spectrum disorders and autoimmune encephalopathies. Frontal gliomas present with apathy and disinhibition, and temporal tumors lead to hallucinations, emotional lability, and episodic memory dysfunction. Tumor-associated neuropsychiatric dysfunction, despite increasing recognition, is underdiagnosed and commonly misdiagnosed. This paper seeks to consolidate the mechanistic understanding of these syndromes, drawing on perspectives from neuroimaging, molecular oncology, neuroimmunology, and computational psychiatry. Novel approaches, including lesion-network mapping, exosomal biomarkers or AI-based predictive modeling, have projected early detection and precision-targeted interventions. In the context of the limitations of conventional psychotropic treatments, mechanistically informed therapies, including neuromodulation, neuroimmune-based interventions, and metabolic reprogramming, are essential to improving psychiatric and oncological outcomes. Paraneoplastic neuropsychiatric syndromes are not due to a secondary effect, rather, they are manifestations integral to the biology of a tumor, so they require a new paradigm in both diagnosis and treatment. And defining their molecular and circuit-level underpinnings will propel the next frontier of precision psychiatry in neuro-oncology, cementing the understanding that psychiatric dysfunction is a core influencer of survival, resilience, and quality of life.