Answers are grounded in this paper and cite their sources.
Background Major depressive disorder (MDD) frequently complicates the clinical course of diffuse large B-cell lymphoma (DLBCL). This study aimed to develop machine learning models to predict new-onset MDD during treatment and to i explore hypothesis-generating neuro-immune predictive associations using SHAP analysis. Methods A retrospective cohort of 526 DLBCL patients was analyzed. Fifteen baseline variables were utilized to train and evaluate six predictive algorithms. SHAP dependence and individual conditional expectation analyses were applied to explore multidimensional feature interactions. Results The Random Forest model demonstrated superior predictive performance in the testing set (AUC = 0.913, 95% CI: 0.858-0.968; Brier score = 0.124). SHAP analysis indicated that constitutional symptom burden (B symptoms), systemic inflammation (IL-6), baseline subclinical depressive symptoms (PHQ-9 score), tumor burden (LDH), and endocrine stress (morning cortisol) were the top five predictive features associated with model-estimated MDD risk. Exploratory dependence and interaction analyses suggested non-linear inflection patterns and statistical interactions, where concurrent elevations in IL-6 and cortisol were associated with higher predicted probabilities. Conclusions Machine learning approaches may facilitate early risk estimation for MDD in DLBCL patients. The identified neuro-immune associations warrant further prospective validation to evaluate their clinical utility.
No takes yet. Share an insight, caveat, or question.
Yan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: