AI-ENABLED PATIENT-DERIVED NEURONAL MODELS FOR MECHANISM-GUIDED ANTIDEPRESSANT SELECTION IN MAJOR DEPRESSIVE DISORDER

Background

Major depressive disorder (MDD) is characterized by dysregulated synaptic plasticity, yet antidepressant selection remains largely empirical, contributing to prolonged trial-and-error treatment and poor outcomes. Artificial intelligence offers powerful tools to quantify complex biological phenotypes, but its clinical value depends on grounding algorithms in mechanistic, patient-specific biology. Translational models that link AI-enabled neuronal readouts to real-world treatment response are needed to define when and for whom AI can meaningfully support clinical decision-making in psychiatry.

Methods

Human induced pluripotent stem cells (hiPSCs) were generated from participants in the STAR*D study and differentiated into cortical excitatory neurons. Patients were stratified based on clinical antidepressant response using QIDS criteria ( > 70% improvement vs < 30%). Neurons were treated in vitro with the same antidepressants received clinically. Synaptic plasticity was quantified using AI-enabled high-content imaging, combining automated dendritic spine detection, connectivity analysis, and pattern recognition to generate objective, scalable measures of neuroplasticity across patient-derived neuronal cultures.

Results

Assay sensitivity and biological validity were first confirmed using brain-derived neurotrophic factor (BDNF), which produced a robust, dose-dependent increase in synaptic plasticity across neurons derived from 16 patients (F = 112.58, p < 0.01). Following validation, antidepressant exposure induced significantly greater synaptic plasticity in neurons derived from clinical responders compared with non-responders. The magnitude and direction of neuronal responses closely mirrored individual STAR*D clinical outcomes and were consistent across antidepressant classes.

Conclusions

AI-enabled analysis of patient-derived neuronal synaptic plasticity provides a mechanistic, functional correlate of antidepressant response, offering a biologically grounded approach to precision psychopharmacology. These findings illustrate how AI can enhance, not replace, clinical judgment by translating complex cellular phenotypes into clinically relevant decision support signals.