Elucidating depression heterogeneity using clinical, neuroimaging and genetic data
Appears in the following Collection
- Psykologisk institutt 
AbstractDepression is a debilitating disorder with a high prevalence compared to most other mental disorders. It has become increasingly apparent that depression is clinically heterogenous, most notably in terms of the breadth and range of possible symptom configurations and profiles, which likely plays into the lack of robust neuroimaging markers. The complex nature of depression is also reflected in its genetic architecture, and comorbidity with other mental disorders, in particular with anxiety, and showing strong links and phenomenological overlap with neuroticism. The main aim of this thesis is to elucidate depression heterogeneity through symptoms, advanced neuroimaging and genetics, which hopefully will further our understanding of the complex phenomenology and etiology of depression. Our approach in paper one was to identify subgroups of depression based on symptoms of anxiety and depression using a data-driven clustering approach, across healthy controls and patients, and to test to which extent the symptom-based subtyping was supported by functional imaging. This approach yielded five subgroups of depression characterized by specific symptom profiles as assessed with centrality measures from network-based graph theory, with a presence of cases and controls across all subgroups. These subgroups were supported by differences in brain functional connectivity patterns from a subsample with resting-state functional magnetic resonance imaging (fMRI), which in particular implicated a fronto-temporal brain network. In contrast, we found no significant associations between brain functional connectivity patterns and dimensional or categorical measures of depression. To make inferences at the individual level, we used a machine learning approach in paper II to map various conceptualizations of resting-state fMRI-based brain functional connectivity to cognitive and mental health traits related to depression in a large population-based cohort (UK Biobank). We also predicted age and classified sex to use as a benchmark comparison for our predictions. Further, we wanted to assess the mapping between the genetic underpinning of these and related traits with brain functional connectivity using polygenic risk scores based on previously published genome-wise association studies. Our results showed high prediction accuracy for age and sex, as well as robust prediction of fluid intelligence and years of educational attainment. In contrast, we observed low prediction accuracy for symptom loads of depression and anxiety and trait level neuroticism, as well as all of the polygenic risk scores. Another explanation for the inconsistent findings in the neuroimaging literature of depression is the relative absence of studies integrating different imaging modalities together. To this end, we used linked independent component analysis to combine data from morphometric, diffusion weighted and resting-state fMRI in individuals with or without a history of depression. Our findings revealed strong associations with age and sex with brain components related to global properties of cortical macrostructure, diffusion-based properties of white matter integrity, and default mode network amplitude. In contrast, we found no association of these brain components with case-control status (depression vs healthy controls), nor symptom loads for depression and anxiety across groups, or any interaction effects with age or sex. The machine learning analyses were generally in line with the univariate analyses, showing low model performance for classifying cases from controls and predicting symptom loads for depression and anxiety, but high model performance for predicting age. The findings of the three papers and thesis confirms the clinical heterogeneity of depression, and that advanced neuroimaging by itself is not enough to elucidate the neurobiological underpinnings. One of the keys to solving the depression riddle are more precise methods of stratification at the individual level. This could ultimately pave a path towards the development of personalized treatment and prevention of depression.
List of papers
|Paper I: Maglanoc LA, Landrø NI, Jonassen R, Kaufmann T, Córdova-Palomera A, Hilland E, & Westlye LT. Data-driven clustering reveals a link between symptoms and functional brain connectivity in depression. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging. 2019;4(1):16-26. DOI: 10.1016/j.bpsc.2018.05.005. The article is not available in DUO due to publisher restrictions. The published version is available at: https://doi.org/10.1016/j.bpsc.2018.05.005|
|Paper II: Maglanoc LA, Kaufmann T, Van der Meer D, Marquand AF, Wolfers T, Jonassen R, Hilland E, Andreassen OA, Landrø NI, & Westlye LT. Brain connectome mapping of complex human traits and their polygenic architecture using machine learning. Biological Psychiatry, 2019. DOI: 10.1016/j.biopsych.2019.10.011. The paper is included in the thesis. The published version is available at: https://doi.org/10.1016/j.biopsych.2019.10.011|
|Paper III: Maglanoc LA, Kaufmann T, Jonassen R, Hilland E, Beck D, Landrø NI, & Westlye LT. Multimodal fusion of structural and functional brain imaging in depression using linked independent component analysis. Human Brain Mapping, 2020; 41: 241– 255. DOI: 10.1002/hbm.24802. The paper is included in the thesis. The published version is available at: https://doi.org/10.1002/hbm.24802|