ABSTRACT
Managing psychiatric disorders, including depression, anxiety, and bipolar disorder, during pregnancy presents significant clinical challenges due to uncertainties surrounding medication safety and efficacy for both the mother and fetus. This study introduces a novel deep learning (DL)-based decision-support framework aimed at personalizing pharmacotherapy for pregnant patients diagnosed with psychiatric conditions. By leveraging electronic health records (EHRs), pharmacogenomic data, and advanced machine learning techniques, we developed a predictive neural network model capable of recommending precise drug classes and dosages tailored to individual patient profiles. Our methodology consisted of clearly defined sequential phases: problem definition, data collection, preprocessing, feature engineering, exploratory data analysis (EDA), model development, genetic data integration, validation, and deployment into clinical practice. Exploratory analysis revealed critical insights, identifying significant predictors of medication efficacy and side effects through visualizations including pair plots, correlation heatmaps, violin plots, and interactive scatter plots. The developed neural network model, optimized using rigorous hyperparameter tuning, exhibited high accuracy and robust predictive power, as evidenced by outstanding ROC-AUC and precision-recall performance metrics. Integration of pharmacogenomic data further improved predictive accuracy, demonstrating the model’s ability to capture complex genetic interactions influencing drug response variability. Rigorous validation through retrospective cohort studies confirmed the clinical applicability and reliability of our system. Ultimately, our decision-support tool provides clinicians with evidence-based, individualized medication strategies that effectively balance therapeutic outcomes and safety concerns during pregnancy. The application of this DL-driven personalized pharmacotherapy framework holds significant potential for enhancing clinical decision-making, improving maternal mental health, and minimizing risks to fetal development.