708. From molecules to meaning: metabolomic and transcriptomic pathways to overcoming treatment resistance

Abstract

Abstract Background Treatment resistance remains a major barrier to improving outcomes in severe psychiatric disorders, reflecting substantial biological heterogeneity that is not captured by symptom-based diagnosis alone. Emerging multi-omics approaches—particularly metabolomic profiling combined with transcriptome-informed network analyses—offer a mechanistic route to identify molecular signatures of treatment responsiveness and resistance and to translate them into actionable stratification tools. Aims & Objectives This abstract summarizes recent work by Prof. Dr. Bernhard Baune focusing on (i) elucidating molecular mechanisms that distinguish treatment-resistant from treatment-responsive patients, (ii) integrating metabolomic and transcriptomic network signals to build biologically grounded predictors of outcome, and (iii) evaluating whether such predictors can inform precision treatment selection across pharmacological strategies, including rapid-acting antidepressant paradigms and anti-inflammatory augmentation. Method Across randomized controlled trials and well-characterized clinical cohorts, metabolomic profiles were quantified from peripheral biospecimens and integrated with transcriptome-derived network features (e.g., co-expression modules and pathway-level activity scores). Illustrative datasets included RCTs contrasting ketamine with midazolam and trials evaluating celecoxib augmentation. Multi-layer integration used supervised machine learning for outcome prediction (response/remission trajectories, symptom change), with cross-validation and external validation where available. Network-based analyses prioritized interpretable biological mechanisms (immune-metabolic signaling, mitochondrial energetics, oxidative stress, and kynurenine-related pathways), linking molecular patterns to clinical endpoints. Results Integrative models combining metabolomic features with transcriptome-based network metrics consistently outperformed single-layer approaches in predicting treatment outcomes. Distinct molecular signatures separated responders from non-responders, including metabolite patterns suggestive of altered inflammatory tone and energy metabolism, alongside transcriptomic modules implicating immune activation and synaptic plasticity pathways. Notably, metabolomic stratification improved prediction even within a priori treatment-resistant samples, indicating that resistance is biologically heterogeneous and partially reversible when mechanistically targeted. In augmentation contexts, inflammatory/metabolic signatures aligned with differential benefit from anti-inflammatory add-on strategies. Discussion & Conclusions These findings support a clinically relevant bridge between molecular biology and treatment efficacy: multi-omics signatures can (i) clarify mechanistic drivers of resistance, (ii) enable patient stratification beyond diagnosis, and (iii) provide predictive biomarkers suitable for decision support. Future work should focus on prospective, multi-site validation, harmonized metabolomics pipelines, and implementation-ready clinical decision models to accelerate precision psychiatry in routine care.