Peripheral molecular signatures of major depressive disorder: a systematic review of blood-derived biomarkers for diagnosis, stratification, and treatment response

Major depressive disorder (MDD) is biologically heterogeneous, yet diagnosis and treatment selection remain largely symptom-based. Blood-derived molecular biomarkers may offer scalable tools for biomarker discovery, biological stratification, and treatment-response research, but their clinical readiness remains uncertain. This PRISMA-compliant systematic review was registered in PROSPERO (CRD420261379864). PubMed, Scopus, and Embase were searched from 2014 to 27 April 2026 for human studies evaluating blood-derived molecular biomarkers in MDD or clinically defined depressive phenotypes with extractable depression-relevant data. Eligible domains included transcriptomics, DNA methylation, non-coding RNA, immune-inflammatory, mitochondrial/metabolic, neurotrophic/neuroendocrine, multi-omics, and machine-learning biomarker studies. Owing to methodological and platform heterogeneity, a structured narrative synthesis was performed rather than quantitative meta-analysis. Methodological quality was assessed using a modified QUADOMICS framework, prediction-model studies were additionally assessed using PROBAST, and certainty was judged using a modified GRADE approach. Of 9,354 records identified, 25 reports met the eligibility criteria, representing 24 independent participant datasets because two methylation analyses used the same underlying NESDA cohort. Included studies used whole blood, PBMCs, leukocytes, monocytes, serum, plasma, and cell-free RNA across diverse molecular platforms. Individual biomarkers were heterogeneous, but recurrent pathway-level signals involved immune-inflammatory and interferon-related pathways, NF-κB signalling, mitochondrial and energy metabolism, DNA methylation and epigenetic regulation, neurodevelopmental/synaptic pathways, stress-response signalling, autophagy, and non-coding RNA regulation. Multi-marker and multi-omics panels showed promising apparent diagnostic or treatment-response relevance, but most were internally validated, cohort-specific, or exploratory. Modified GRADE certainty was low to very low across biomarker domains. Blood-derived molecular signatures in MDD show stronger convergence at the pathway level than at the individual-biomarker level. Current evidence supports hypothesis generation and biological stratification, but no blood-derived biomarker or panel is ready for routine diagnosis or treatment selection.