The circulating proteome and brain health: Mendelian randomisation and cross-sectional analyses

Decline in cognitive function is the most feared aspect of ageing. Poorer midlife cognitive function is associated with increased dementia and stroke risk. The mechanisms underlying variation in cognitive function are uncertain. Here, we assessed associations between 1160 proteins' plasma levels and two measures of cognitive function, the digit symbol substitution test (DSST) and the Montreal Cognitive Assessment in 1198 PURE-MIND participants. We identified five DSST performance-associated proteins (NCAN, BCAN, CA14, MOG, CDCP1), with NCAN and CDCP1 showing replicated association in an independent cohort, GS (N = 1053). MRI-assessed structural brain phenotypes partially mediated (8-19%) associations between NCAN, BCAN, and MOG, and DSST performance. Mendelian randomisation analyses suggested higher CA14 levels might cause larger hippocampal volume and increased stroke risk, whilst higher CDCP1 levels might increase intracranial aneurysm risk. Our findings highlight candidates for further study and the potential for drug repurposing to reduce the risk of stroke and cognitive decline.

Disrupted limbic-prefrontal effective connectivity in response to fearful faces in lifetime depression

BACKGROUND

Multiple brain imaging studies of negative emotional bias in major depressive disorder (MDD) have used images of fearful facial expressions and focused on the amygdala and the prefrontal cortex. The results have, however, been inconsistent, potentially due to small sample sizes (typically N<50). It remains unclear if any alterations are a characteristic of current depression or of past experience of depression, and whether there are MDD-related changes in effective connectivity between the two brain regions.

METHODS

Activations and effective connectivity between the amygdala and dorsolateral prefrontal cortex (DLPFC) in response to fearful face stimuli were studied in a large population-based sample from Generation Scotland. Participants either had no history of MDD (N=664 in activation analyses, N=474 in connectivity analyses) or had a diagnosis of MDD during their lifetime (LMDD, N=290 in activation analyses, N=214 in connectivity analyses). The within-scanner task involved implicit facial emotion processing of neutral and fearful faces.

RESULTS

Compared to controls, LMDD was associated with increased activations in left amygdala (PFWE=0.031,kE=4) and left DLPFC (PFWE=0.002,kE=33), increased mean bilateral amygdala activation (β=0.0715,P=0.0314), and increased inhibition from left amygdala to left DLPFC, all in response to fearful faces contrasted to baseline. Results did not appear to be attributable to depressive illness severity or antidepressant medication status at scan time.

LIMITATIONS

Most studied participants had past rather than current depression, average severity of ongoing depression symptoms was low, and a substantial proportion of participants were receiving medication. The study was not longitudinal and the participants were only assessed a single time.

CONCLUSIONS

LMDD is associated with hyperactivity of the amygdala and DLPFC, and with stronger amygdala to DLPFC inhibitory connectivity, all in response to fearful faces, unrelated to depression severity at scan time. These results help reduce inconsistency in past literature and suggest disruption of 'bottom-up' limbic-prefrontal effective connectivity in depression.

A comprehensive hierarchical comparison of structural connectomes in Major Depressive Disorder cases v. controls in two large population samples

BACKGROUND

The brain can be represented as a network, with nodes as brain regions and edges as region-to-region connections. Nodes with the most connections (hubs) are central to efficient brain function. Current findings on structural differences in Major Depressive Disorder (MDD) identified using network approaches remain inconsistent, potentially due to small sample sizes. It is still uncertain at what level of the connectome hierarchy differences may exist, and whether they are concentrated in hubs, disrupting fundamental brain connectivity.

METHODS

We utilized two large cohorts, UK Biobank (UKB, N = 5104) and Generation Scotland (GS, N = 725), to investigate MDD case-control differences in brain network properties. Network analysis was done across four hierarchical levels: (1) global, (2) tier (nodes grouped into four tiers based on degree) and rich club (between-hub connections), (3) nodal, and (4) connection.

RESULTS

In UKB, reductions in network efficiency were observed in MDD cases globally (d = -0.076, pFDR = 0.033), across all tiers (d = -0.069 to -0.079, pFDR = 0.020), and in hubs (d = -0.080 to -0.113, pFDR = 0.013-0.035). No differences in rich club organization and region-to-region connections were identified. The effect sizes and direction for these associations were generally consistent in GS, albeit not significant in our lower-N replication sample.

CONCLUSION

Our results suggest that the brain's fundamental rich club structure is similar in MDD cases and controls, but subtle topological differences exist across the brain. Consistent with recent large-scale neuroimaging findings, our findings offer a connectomic perspective on a similar scale and support the idea that minimal differences exist between MDD cases and controls.

General and specific patterns of cortical gene expression as spatial correlates of complex cognitive functioning

Gene expression varies across the brain. This spatial patterning denotes specialised support for particular brain functions. However, the way that a given gene's expression fluctuates across the brain may be governed by general rules. Quantifying patterns of spatial covariation across genes would offer insights into the molecular characteristics of brain areas supporting, for example, complex cognitive functions. Here, we use principal component analysis to separate general and unique gene regulatory associations with cortical substrates of cognition. We find that the region-to-region variation in cortical expression profiles of 8235 genes covaries across two major principal components: gene ontology analysis suggests these dimensions are characterised by downregulation and upregulation of cell-signalling/modification and transcription factors. We validate these patterns out-of-sample and across different data processing choices. Brain regions more strongly implicated in general cognitive functioning (g; 3 cohorts, total meta-analytic N = 39,519) tend to be more balanced between downregulation and upregulation of both major components (indicated by regional component scores). We then identify a further 29 genes as candidate cortical spatial correlates of g, beyond the patterning of the two major components (|β| range = 0.18 to 0.53). Many of these genes have been previously associated with clinical neurodegenerative and psychiatric disorders, or with other health-related phenotypes. The results provide insights into the cortical organisation of gene expression and its association with individual differences in cognitive functioning.

Comprehensive assessment of sleep duration, insomnia, and brain structure within the UK Biobank cohort

STUDY OBJECTIVES

To assess for associations between sleeping more than or less than recommended by the National Sleep Foundation (NSF), and self-reported insomnia, with brain structure.

METHODS

Data from the UK Biobank cohort were analyzed (N between 9K and 32K, dependent on availability, aged 44 to 82 years). Sleep measures included self-reported adherence to NSF guidelines on sleep duration (sleeping between 7 and 9 hours per night), and self-reported difficulty falling or staying asleep (insomnia). Brain structural measures included global and regional cortical or subcortical morphometry (thickness, surface area, volume), global and tract-related white matter microstructure, brain age gap (difference between chronological age and age estimated from brain scan), and total volume of white matter lesions.

RESULTS

Longer-than-recommended sleep duration was associated with lower overall grey and white matter volumes, lower global and regional cortical thickness and volume measures, higher brain age gap, higher volume of white matter lesions, higher mean diffusivity globally and in thalamic and association fibers, and lower volume of the hippocampus. Shorter-than-recommended sleep duration was related to higher global and cerebellar white matter volumes, lower global and regional cortical surface areas, and lower fractional anisotropy in projection fibers. Self-reported insomnia was associated with higher global gray and white matter volumes, and with higher volumes of the amygdala, hippocampus, and putamen.

CONCLUSIONS

Sleeping longer than recommended by the NSF is associated with a wide range of differences in brain structure, potentially indicative of poorer brain health. Sleeping less than recommended is distinctly associated with lower cortical surface areas. Future studies should assess the potential mechanisms of these differences and investigate long sleep duration as a putative marker of brain health.

General and specific patterns of cortical gene expression as spatial correlates of complex cognitive functioning

Gene expression varies across the brain. This spatial patterning denotes specialised support for particular brain functions. However, the way that a given gene's expression fluctuates across the brain may be governed by general rules. Quantifying patterns of spatial covariation across genes would offer insights into the molecular characteristics of brain areas supporting, for example, complex cognitive functions. Here, we use principal component analysis to separate general and unique gene regulatory associations with cortical substrates of cognition. We find that the region-to-region variation in cortical expression profiles of 8235 genes covaries across two major principal components : gene ontology analysis suggests these dimensions are characterised by downregulation and upregulation of cell-signalling/modification and transcription factors. We validate these patterns out-of-sample and across different data processing choices. Brain regions more strongly implicated in general cognitive functioning (g; 3 cohorts, total meta-analytic N = 39,519) tend to be more balanced between downregulation and upregulation of both major components (indicated by regional component scores). We then identify a further 41 genes as candidate cortical spatial correlates of g, beyond the patterning of the two major components (|β| range = 0.15 to 0.53). Many of these genes have been previously associated with clinical neurodegenerative and psychiatric disorders, or with other health-related phenotypes. The results provide insights into the cortical organisation of gene expression and its association with individual differences in cognitive functioning.

Predicting sex, age, general cognition and mental health with machine learning on brain structural connectomes

There is an increasing expectation that advanced, computationally expensive machine learning (ML) techniques, when applied to large population-wide neuroimaging datasets, will help to uncover key differences in the human brain in health and disease. We take a comprehensive approach to explore how multiple aspects of brain structural connectivity can predict sex, age, general cognitive function and general psychopathology, testing different ML algorithms from deep learning (DL) model (BrainNetCNN) to classical ML methods. We modelled N = 8183 structural connectomes from UK Biobank using six different structural network weightings obtained from diffusion MRI. Streamline count generally provided the highest prediction accuracies in all prediction tasks. DL did not improve on prediction accuracies from simpler linear models. Further, high correlations between gradient attribution coefficients from DL and model coefficients from linear models suggested the models ranked the importance of features in similar ways, which indirectly suggested the similarity in models' strategies for making predictive decision to some extent. This highlights that model complexity is unlikely to improve detection of associations between structural connectomes and complex phenotypes with the current sample size.

AI-based dimensional neuroimaging system for characterizing heterogeneity in brain structure and function in major depressive disorder: COORDINATE-MDD consortium design and rationale

BACKGROUND

Efforts to develop neuroimaging-based biomarkers in major depressive disorder (MDD), at the individual level, have been limited to date. As diagnostic criteria are currently symptom-based, MDD is conceptualized as a disorder rather than a disease with a known etiology; further, neural measures are often confounded by medication status and heterogeneous symptom states.

METHODS

We describe a consortium to quantify neuroanatomical and neurofunctional heterogeneity via the dimensions of novel multivariate coordinate system (COORDINATE-MDD). Utilizing imaging harmonization and machine learning methods in a large cohort of medication-free, deeply phenotyped MDD participants, patterns of brain alteration are defined in replicable and neurobiologically-based dimensions and offer the potential to predict treatment response at the individual level. International datasets are being shared from multi-ethnic community populations, first episode and recurrent MDD, which are medication-free, in a current depressive episode with prospective longitudinal treatment outcomes and in remission. Neuroimaging data consist of de-identified, individual, structural MRI and resting-state functional MRI with additional positron emission tomography (PET) data at specific sites. State-of-the-art analytic methods include automated image processing for extraction of anatomical and functional imaging variables, statistical harmonization of imaging variables to account for site and scanner variations, and semi-supervised machine learning methods that identify dominant patterns associated with MDD from neural structure and function in healthy participants.

RESULTS

We are applying an iterative process by defining the neural dimensions that characterise deeply phenotyped samples and then testing the dimensions in novel samples to assess specificity and reliability. Crucially, we aim to use machine learning methods to identify novel predictors of treatment response based on prospective longitudinal treatment outcome data, and we can externally validate the dimensions in fully independent sites.

CONCLUSION

We describe the consortium, imaging protocols and analytics using preliminary results. Our findings thus far demonstrate how datasets across many sites can be harmonized and constructively pooled to enable execution of this large-scale project.

Virtual Ontogeny of Cortical Growth Preceding Mental Illness

BACKGROUND

Morphology of the human cerebral cortex differs across psychiatric disorders, with neurobiology and developmental origins mostly undetermined. Deviations in the tangential growth of the cerebral cortex during pre/perinatal periods may be reflected in individual variations in cortical surface area later in life.

METHODS

Interregional profiles of group differences in surface area between cases and controls were generated using T1-weighted magnetic resonance imaging from 27,359 individuals including those with attention-deficit/hyperactivity disorder, autism spectrum disorder, bipolar disorder, major depressive disorder, schizophrenia, and high general psychopathology (through the Child Behavior Checklist). Similarity of interregional profiles of group differences in surface area and prenatal cell-specific gene expression was assessed.

RESULTS

Across the 11 cortical regions, group differences in cortical area for attention-deficit/hyperactivity disorder, schizophrenia, and Child Behavior Checklist were dominant in multimodal association cortices. The same interregional profiles were also associated with interregional profiles of (prenatal) gene expression specific to proliferative cells, namely radial glia and intermediate progenitor cells (greater expression, larger difference), as well as differentiated cells, namely excitatory neurons and endothelial and mural cells (greater expression, smaller difference). Finally, these cell types were implicated in known pre/perinatal risk factors for psychosis. Genes coexpressed with radial glia were enriched with genes implicated in congenital abnormalities, birth weight, hypoxia, and starvation. Genes coexpressed with endothelial and mural genes were enriched with genes associated with maternal hypertension and preterm birth.

CONCLUSIONS

Our findings support a neurodevelopmental model of vulnerability to mental illness whereby prenatal risk factors acting through cell-specific processes lead to deviations from typical brain development during pregnancy.

Integrated methylome and phenome study of the circulating proteome reveals markers pertinent to brain health

Characterising associations between the methylome, proteome and phenome may provide insight into biological pathways governing brain health. Here, we report an integrated DNA methylation and phenotypic study of the circulating proteome in relation to brain health. Methylome-wide association studies of 4058 plasma proteins are performed (N = 774), identifying 2928 CpG-protein associations after adjustment for multiple testing. These are independent of known genetic protein quantitative trait loci (pQTLs) and common lifestyle effects. Phenome-wide association studies of each protein are then performed in relation to 15 neurological traits (N = 1,065), identifying 405 associations between the levels of 191 proteins and cognitive scores, brain imaging measures or APOE e4 status. We uncover 35 previously unreported DNA methylation signatures for 17 protein markers of brain health. The epigenetic and proteomic markers we identify are pertinent to understanding and stratifying brain health.

Prediction of depression symptoms in individual subjects with face and eye movement tracking

BACKGROUND

Depression is a challenge to diagnose reliably and the current gold standard for trials of DSM-5 has been in agreement between two or more medical specialists. Research studies aiming to objectively predict depression have typically used brain scanning. Less expensive methods from cognitive neuroscience may allow quicker and more reliable diagnoses, and contribute to reducing the costs of managing the condition. In the current study we aimed to develop a novel inexpensive system for detecting elevated symptoms of depression based on tracking face and eye movements during the performance of cognitive tasks.

METHODS

In total, 75 participants performed two novel cognitive tasks with verbal affective distraction elements while their face and eye movements were recorded using inexpensive cameras. Data from 48 participants (mean age 25.5 years, standard deviation of 6.1 years, 25 with elevated symptoms of depression) passed quality control and were included in a case-control classification analysis with machine learning.

RESULTS

Classification accuracy using cross-validation (within-study replication) reached 79% (sensitivity 76%, specificity 82%), when face and eye movement measures were combined. Symptomatic participants were characterised by less intense mouth and eyelid movements during different stages of the two tasks, and by differences in frequencies and durations of fixations on affectively salient distraction words.

CONCLUSIONS

Elevated symptoms of depression can be detected with face and eye movement tracking during the cognitive performance, with a close to clinically-relevant accuracy (~80%). Future studies should validate these results in larger samples and in clinical populations.

Brain charts for the human lifespan

Over the past few decades, neuroimaging has become a ubiquitous tool in basic research and clinical studies of the human brain. However, no reference standards currently exist to quantify individual differences in neuroimaging metrics over time, in contrast to growth charts for anthropometric traits such as height and weight1. Here we assemble an interactive open resource to benchmark brain morphology derived from any current or future sample of MRI data ( http://www.brainchart.io/ ). With the goal of basing these reference charts on the largest and most inclusive dataset available, acknowledging limitations due to known biases of MRI studies relative to the diversity of the global population, we aggregated 123,984 MRI scans, across more than 100 primary studies, from 101,457 human participants between 115 days post-conception to 100 years of age. MRI metrics were quantified by centile scores, relative to non-linear trajectories2 of brain structural changes, and rates of change, over the lifespan. Brain charts identified previously unreported neurodevelopmental milestones3, showed high stability of individuals across longitudinal assessments, and demonstrated robustness to technical and methodological differences between primary studies. Centile scores showed increased heritability compared with non-centiled MRI phenotypes, and provided a standardized measure of atypical brain structure that revealed patterns of neuroanatomical variation across neurological and psychiatric disorders. In summary, brain charts are an essential step towards robust quantification of individual variation benchmarked to normative trajectories in multiple, commonly used neuroimaging phenotypes.

Comparing personalized brain-based and genetic risk scores for major depressive disorder in large population samples of adults and adolescents

Background

Major depressive disorder (MDD) is a polygenic disorder associated with brain alterations but until recently, there have been no brain-based metrics to quantify individual-level variation in brain morphology. Here, we evaluated and compared the performance of a new brain-based ‘Regional Vulnerability Index’ (RVI) with polygenic risk scores (PRS), in the context of MDD. We assessed associations with syndromal MDD in an adult sample (N = 702, age = 59 ± 10) and with subclinical depressive symptoms in a longitudinal adolescent sample (baseline N = 3,825, age = 10 ± 1; 2-year follow-up N = 2,081, age = 12 ± 1).

Methods

MDD-RVIs quantify the correlation of the individual’s corresponding brain metric with the expected pattern for MDD derived in an independent sample. Using the same methodology across samples, subject-specific MDD-PRS and six MDD-RVIs based on different brain modalities (subcortical volume, cortical thickness, cortical surface area, mean diffusivity, fractional anisotropy, and multimodal) were computed.

Results

In adults, MDD-RVIs (based on white matter and multimodal measures) were more strongly associated with MDD (β = 0.099–0.281, P_FDR = 0.001–0.043) than MDD-PRS (β = 0.056–0.152, P_FDR = 0.140–0.140). In adolescents, depressive symptoms were associated with MDD-PRS at baseline and follow-up (β = 0.084–0.086, p = 1.38 × 10⁻⁴−4.77 × 10⁻⁴) but not with any MDD-RVIs (β < 0.05, p > 0.05).

Conclusions

Our results potentially indicate the ability of brain-based risk scores to capture a broader range of risk exposures than genetic risk scores in adults and are also useful in helping us to understand the temporal origins of depression-related brain features. Longitudinal data, specific to the developmental period and on white matter measures, will be useful in informing risk for subsequent psychiatric illness.

Early life predictors of late life cerebral small vessel disease in four prospective cohort studies

Development of cerebral small vessel disease, a major cause of stroke and dementia, may be influenced by early life factors. It is unclear whether these relationships are independent of each other, of adult socio-economic status or of vascular risk factor exposures. We examined associations between factors from birth (ponderal index, birth weight), childhood (IQ, education, socio-economic status), adult small vessel disease, and brain volumes, using data from four prospective cohort studies: STratifying Resilience And Depression Longitudinally (STRADL) (n = 1080; mean age = 59 years); the Dutch Famine Birth Cohort (n = 118; mean age = 68 years); the Lothian Birth Cohort 1936 (LBC1936; n = 617; mean age = 73 years), and the Simpson's cohort (n = 110; mean age = 78 years). We analysed each small vessel disease feature individually and summed to give a total small vessel disease score (range 1-4) in each cohort separately, then in meta-analysis, adjusted for vascular risk factors and adult socio-economic status. Higher birth weight was associated with fewer lacunes [odds ratio (OR) per 100 g = 0.93, 95% confidence interval (CI) = 0.88 to 0.99], fewer infarcts (OR = 0.94, 95% CI = 0.89 to 0.99), and fewer perivascular spaces (OR = 0.95, 95% CI = 0.91 to 0.99). Higher childhood IQ was associated with lower white matter hyperintensity burden (OR per IQ point = 0.99, 95% CI 0.98 to 0.998), fewer infarcts (OR = 0.98, 95% CI = 0.97 to 0.998), fewer lacunes (OR = 0.98, 95% CI = 0.97 to 0.999), and lower total small vessel disease burden (OR = 0.98, 95% CI = 0.96 to 0.999). Low education was associated with more microbleeds (OR = 1.90, 95% CI = 1.33 to 2.72) and lower total brain volume (mean difference = -178.86 cm3, 95% CI = -325.07 to -32.66). Low childhood socio-economic status was associated with fewer lacunes (OR = 0.62, 95% CI = 0.40 to 0.95). Early life factors are associated with worse small vessel disease in later life, independent of each other, vascular risk factors and adult socio-economic status. Risk for small vessel disease may originate in early life and provide a mechanistic link between early life factors and risk of stroke and dementia. Policies investing in early child development may improve lifelong brain health and contribute to the prevention of dementia and stroke in older age.

Hair glucocorticoids are associated with childhood adversity, depressive symptoms and reduced global and lobar grey matter in Generation Scotland

Hypothalamic-pituitary-adrenal (HPA) axis dysregulation has been commonly reported in major depressive disorder (MDD), but with considerable heterogeneity of results; potentially due to the predominant use of acute measures of an inherently variable/phasic system. Chronic longer-term measures of HPA-axis activity have yet to be systematically examined in MDD, particularly in relation to brain phenotypes, and in the context of early-life/contemporaneous stress. Here, we utilise a temporally stable measure of cumulative HPA-axis function (hair glucocorticoids) to investigate associations between cortisol, cortisone and total glucocorticoids with concurrent measures of (i) lifetime-MDD case/control status and current symptom severity, (ii) early/current-life stress and (iii) structural neuroimaging phenotypes, in N = 993 individuals from Generation Scotland (mean age = 59.1 yrs). Increased levels of hair cortisol were significantly associated with reduced global and lobar brain volumes with reductions in the frontal, temporal and cingulate regions (βrange = -0.057 to -0.104, all PFDR < 0.05). Increased levels of hair cortisone were significantly associated with MDD (lifetime-MDD status, current symptoms, and severity; βrange = 0.071 to 0.115, all PFDR = < 0.05), with early-life adversity (β = 0.083, P = 0.017), and with reduced global and regional brain volumes (global: β = -0.059, P = 0.043; nucleus accumbens: β = -0.075, PFDR = 0.044). Associations with total glucocorticoids followed a similar pattern to the cortisol findings. In this large community-based sample, elevated glucocorticoids were significantly associated with MDD, with early, but not later-life stress, and with reduced global and regional brain phenotypes. These findings provide important foundations for future mechanistic studies to formally explore causal relationships between early adversity, chronic rather than acute measures of glucocorticoids, and neurobiological associations relevant to the aetiology of MDD.

Identification of plasma proteins relating to brain neurodegeneration and vascular pathology in cognitively normal individuals

INTRODUCTION

This study aims to first discover plasma proteomic biomarkers relating to neurodegeneration (N) and vascular (V) damage in cognitively normal individuals and second to discover proteins mediating sex-related difference in N and V pathology.

METHODS

Five thousand and thirty-two plasma proteins were measured in 1061 cognitively normal individuals (628 females and 433 males), nearly 90% of whom had magnetic resonance imaging measures of hippocampal volume (as N) and white matter hyperintensities (as V).

RESULTS

Differential protein expression analysis and co-expression network analysis revealed different proteins and modules associated with N and V, respectively. Furthermore, causal mediation analysis revealed four proteins mediated sex-related difference in N and one protein mediated such difference in V damage.

DISCUSSION

Once validated, the identified proteins could help to select cognitively normal individuals with N and V pathology for Alzheimer's disease clinical trials and provide targets for further mechanistic studies on brain sex differences, leading to sex-specific therapeutic strategies.

Spectral clustering based on structural magnetic resonance imaging and its relationship with major depressive disorder and cognitive ability

There is increasing interest in using data-driven unsupervised methods to identify structural underpinnings of common mental illnesses, including major depressive disorder (MDD) and associated traits such as cognition. However, studies are often limited to severe clinical cases with small sample sizes and most do not include replication. Here, we examine two relatively large samples with structural magnetic resonance imaging (MRI), measures of lifetime MDD and cognitive variables: Generation Scotland (GS subsample, N = 980) and UK Biobank (UKB, N = 8,900), for discovery and replication, using an exploratory approach. Regional measures of FreeSurfer derived cortical thickness (CT), cortical surface area (CSA), cortical volume (CV) and subcortical volume (subCV) were input into a clustering process, controlling for common covariates. The main analysis steps involved constructing participant K-nearest neighbour graphs and graph partitioning with Markov stability to determine optimal clustering of participants. Resultant clusters were (1) checked whether they were replicated in an independent cohort and (2) tested for associations with depression status and cognitive measures. Participants separated into two clusters based on structural brain measurements in GS subsample, with large Cohen's d effect sizes between clusters in higher order cortical regions, commonly associated with executive function and decision making. Clustering was replicated in the UKB sample, with high correlations of cluster effect sizes for CT, CSA, CV and subCV between cohorts across regions. The identified clusters were not significantly different with respect to MDD case-control status in either cohort (GS subsample: p_FDR  = .2239-.6585; UKB: p_FDR  = .2003-.7690). Significant differences in general cognitive ability were, however, found between the clusters for both datasets, for CSA, CV and subCV (GS subsample: d = 0.2529-.3490, p_FDR  < .005; UKB: d = 0.0868-0.1070, p_FDR  < .005). Our results suggest that there are replicable natural groupings of participants based on cortical and subcortical brain measures, which may be related to differences in cognitive performance, but not to the MDD case-control status.

Automated classification of depression from structural brain measures across two independent community-based cohorts

Major depressive disorder (MDD) has been the subject of many neuroimaging case-control classification studies. Although some studies report accuracies ≥80%, most have investigated relatively small samples of clinically-ascertained, currently symptomatic cases, and did not attempt replication in larger samples. We here first aimed to replicate previously reported classification accuracies in a small, well-phenotyped community-based group of current MDD cases with clinical interview-based diagnoses (from STratifying Resilience and Depression Longitudinally cohort, 'STRADL'). We performed a set of exploratory predictive classification analyses with measures related to brain morphometry and white matter integrity. We applied three classifier types-SVM, penalised logistic regression or decision tree-either with or without optimisation, and with or without feature selection. We then determined whether similar accuracies could be replicated in a larger independent population-based sample with self-reported current depression (UK Biobank cohort). Additional analyses extended to lifetime MDD diagnoses-remitted MDD in STRADL, and lifetime-experienced MDD in UK Biobank. The highest cross-validation accuracy (75%) was achieved in the initial current MDD sample with a decision tree classifier and cortical surface area features. The most frequently selected decision tree split variables included surface areas of bilateral caudal anterior cingulate, left lingual gyrus, left superior frontal, right precentral and paracentral regions. High accuracy was not achieved in the larger samples with self-reported current depression (53.73%), with remitted MDD (57.48%), or with lifetime-experienced MDD (52.68-60.29%). Our results indicate that high predictive classification accuracies may not immediately translate to larger samples with broader criteria for depression, and may not be robust across different classification approaches.