Ramona Leenings

Background: Despite established consensus about the individuality of brain physiology and structural changes, current neuroimaging biomarker paradigms heavily rely on population averages. Here, we define normativity not as a single reference point but as a landscape of permissible configurations arising from diverse physiological manifestations. Methods: We introduce the Nearest Neighbor Normativity (N ³ ) framework, designed to parse the natural heterogeneity in brain structural measurements into a diversity-aware biomarker of brain structural decline. Through repeated comparisons within different demographic subgroups, N ³ contextualizes an individual’s measurements from several viewpoints along the aging continuum. Relying on local density estimation, it accommodates multiple, shared normative configurations within and across age groups. We use T1-weighted MRI data from 29,883 individuals for training and 7,013 individuals for validation, and benchmark our framework against Brain Age models and Normative Modeling. We evaluate each biomarker’s ability to detect neurodegenerative diseases, using them as representative models of atypical brain structural decline. Results: In group-level analyses, the N ³ biomarker shows the largest effect sizes in the detection of Mild Cognitive Impairment, Alzheimer’s Disease, and Frontotemporal Dementia. Similarly, the N ³ biomarker achieves the highest predictive performance in the classification of neurodegenerative diseases on a single subject level. Notably, the N ³ estimates remain stable across age and robust to compositional changes in the reference sample. Finally, correlation analyses indicate that it captures distinct and complementary aspects of brain structural variation compared to existing approaches. Conclusions: Our findings underscore that rethinking normativity assessments to explicitly accommodate natural inter-individual variability can advance our collective efforts toward truly personalized patient care.

Background: Psychiatric diagnostic practice faces persistent challenges in adequately capturing the complexity, heterogeneity, and dynamic nature of mental disorders. Conventional categorical classification systems such as the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) and the International Classification of Diseases, 11th Revision (ICD-11) rely on rigid symptom-based criteria that often fail to represent multifactorial etiologies, dimensional symptom profiles, and high rates of comorbidity. Although dimensional models, domain-specific ontologies, and alternative taxonomies have been developed to address these limitations, they are typically applied in isolation, resulting in fragmented knowledge representations and limited interoperability. This lack of semantic integration hampers data harmonization and cross-domain analysis. Objective: The objective of this work is to develop and present the Meta Mesh Ontology (MMO) and its Core Domain Framework (CDF) as an integrative approach for mapping heterogeneous psychiatric knowledge bases and evaluating ontology structure, logical consistency, and specified retrieval and inference behavior. Methods: We designed a modular ontology architecture grounded in the Basic Formal Ontology to ensure structural and semantic consistency across biological, psychological, and social domains. Existing psychiatric frameworks and ontologies—including dimensional models, diagnostic taxonomies, and socioenvironmental knowledge systems—were systematically analyzed and represented through domain modules using a documented alignment and mapping procedure. The MMO was implemented using formal semantic web standards, including the Web Ontology Language and the Resource Description Framework (RDF), and aligned with established biomedical ontologies and informatics standards to support explicit mappings and modular extensions across domains. Evaluation used 2 MMO and CDF metric sets and 10 module metric sets to quantify ontology size, semantic alignments, external integration, and dataset readiness. We also assessed Web Ontology Language 2 Description Logic (OWL 2 DL) conformance, logical consistency, and 11 competency questions (including 1 entailment test) and 4 negative controls, in an outpatient scenario with 50 variables. Results: The MMO provides a shared semantic framework for representing links among heterogeneous psychiatric knowledge sources. Its modular design allows domain-specific subontologies to be incorporated under a common architecture. MMO and CDF comprised 2238 triples, 208 classes, 28 properties, 3 individuals, and 103 alignments (100 external) across 5 frameworks. The 10 modules added 967 triples and 171 alignments (116 external), with 21 integrations and complete label coverage (54/54 resources). The ontology resources conformed to OWL 2 DL; HermiT (Information Systems Group) detected no logical inconsistency, and all 11 competency questions and 4 negative controls produced the expected results. All 50 variables were linked to at least 1 CDF anchor, and 47 had an explicit Simple Knowledge Organization System (SKOS) mapping. Conclusions: The MMO provides a modular foundation for representing links among diverse psychiatric knowledge systems and data modalities. The current evaluation characterizes its structure, mappings, and behavior in the specified tests. Further research should assess mapping validity, governed clinical data, operational performance, and clinical utility.

Major depressive disorder (MDD) has been associated with accelerated structural brain aging, yet whether this reflects a pre-existing neurobiological vulnerability, a dynamic acute state effect, or an accumulating biological residual remains unresolved. Across two longitudinal cohorts (N=3220), including a unique sample of 78 initially healthy individuals who transitioned into their first depressive episode during the study course, we systematically tested all three hypotheses. Patients with diagnosed MDD showed elevated MRI-derived brain age relative to healthy controls (1.4 and 2.5 years across cohorts). For the vulnerability hypothesis, individuals scanned prior to their first episode showed no baseline elevation, despite already demonstrating subclinical elevations in self-reported symptom severity, indicating that advanced brain age does not precede illness onset. For the state hypothesis, we found no acceleration of brain aging following the first depressive episode, and longitudinal brain age trajectories were independent of acute clinical symptom severity. Finally, neither episode duration nor recurrence scaled with brain age. Accelerated brain aging in depression is therefore neither an antecedent vulnerability nor an acute state marker of the first episode, but rather a stable biological feature of a long term illness course.

Technological advances in magnetic resonance spectroscopy (MRS), particularly at ultra-high field strengths (e.g., 7T), now allow for the reliable quantification of an increasing number of brain metabolites across multiple brain regions. How-ever, most studies remain constrained to analyses of isolated metabolite levels, overlooking the complex interdependencies that govern neurometabolic regula-tion. Here, we present a multivariate framework that applies network analysis and systematic evaluation of pairwise metabolite ratios to a reference dataset of 53 individuals, revealing coordinated patterns of metabolite covariation as well as a hierarchical inter-metabolite architecture. We validate the generaliz-ability and robustness of these findings across four independent MRS datasets, encompassing both longitudinal and cross-sectional samples from 78 humans and 7 rats across five brain regions and three field strengths. Our results show that inter-metabolite relationships are broadly conserved across methodological contexts, while selective axes exhibit graded, network-informed modulation in response to physiological challenges such as caloric restriction or transcranial magnetic stimulation. By moving beyond univariate measures, our multivariate framework accommodates the innate complexity in human brain metabolism and carries translational relevance for biomarker discovery and the development of metabolically targeted therapeutic strategies.

Objective, scalable biomarkers are needed for continuous monitoring of major depressive disorder (MDD). Smartphone-collected speech is promising, yet extracting clinically useful signals remains difficult. We analysed 3 151 weekly voice diaries from 284 German-speaking adults (128 MDD, 156 controls) and regressed Beck Depression Inventory (BDI) scores. Sentence embeddings from the open-source 8-billion-parameter Qwen3-8B model predicted scores with MAE = 4.45 and R ² = 0.35, explaining 16 more points of variance than the best traditional feature set (TF-IDF). Adding lexical–prosodic or TF-IDF features provided only marginal improvement (best MAE = 4.39). To interpret the embeddings we applied BERTopic and uncovered ten coherent themes; BDI scores peaked for “Persistent Low Mood” and “Pain Distress”, confirming clinical relevance. Large-language-model embeddings therefore capture the dominant signal of depression severity in everyday speech and, paired with interpretable topic analysis, offer a privacy-preserving, scalable route to digital mental-health phenotyping.

Background: Despite established consensus about the individuality of brain physiology and structural changes, current neuroimaging biomarker paradigms heavily rely on population averages. Here, we define normativity not as a single reference point but as a landscape of permissible configurations arising from diverse physiological manifestations. Methods: We introduce the Nearest Neighbor Normativity (N ³ ) framework, designed to parse the natural heterogeneity in brain structural measurements into a diversity-aware biomarker of brain structural decline. Through repeated comparisons within different demographic subgroups, N ³ contextualizes an individual's measurements from several viewpoints along the aging continuum. Relying on local density estimation, it accommodates multiple, shared normative configurations within and across age groups. We use T1-weighted MRI data from 29,883 individuals for training and 7,013 individuals for validation, and benchmark our framework against Brain Age models and Normative Modeling. We evaluate each biomarker's ability to detect neurodegenerative diseases, using them as representative models of atypical brain structural decline. Results: In group-level analyses, the N ³ biomarker shows the largest effect sizes in the detection of Mild Cognitive Impairment, Alzheimer's Disease, and Frontotemporal Dementia. Similarly, the N ³ biomarker achieves the highest predictive performance in the classification of neurodegenerative diseases on a single subject level. Notably, the N ³ estimates remain stable across age and robust to compositional changes in the reference sample. Finally, correlation analyses indicate that it captures distinct and complementary aspects of brain structural variation compared to existing approaches. Conclusions: Our findings underscore that rethinking normativity assessments to explicitly accommodate natural inter-individual variability can enhance the efficacy and translational value of neuroimaging biomarkers and advance our collective efforts toward truly personalized patient care.