Robust pipelines for extensive analyses of large genetic and brain imaging datasets linked to complex human behavior

The rapid technological and methodological advances in genetics, molecular biology and brain imaging in the last decades have enabled the current wide-spread application of brain-wide and genome-wide analyses of potential biological substrates of complex behavioral traits, such as psychological processes and psychiatric disorders. This thesis addresses methodological and statistical issues emerging from the large scale, complexity and explorative nature of brain-wide and genome-wide analyses. Furthermore, it points out additional steps for increasing the confidence in findings resulting from such extensive analyses. It does so by introducing two studies investigating the genetic bases of depressive symptoms and the brain imaging underpinnings of recognition memory performance, respectively. In the first study we aggregated genome-wide data of genetic variation to groups of genes and used inferential statistics to associate them with depressive symptoms. We also replicated the results in an independent sample and used imagining genetics to validate and extend our findings. In the second study, we decomposed the voxel-wise brain activation contrast of looking at previously seen vs. new pictures into 12 brain networks, based of which we evaluated recognition memory performance using prediction analysis. We used stable and reproducible data-driven decomposition and we trained and tested our prediction model in different samples, insuring higher generalizability of our findings. These two studies offer additional insight into the biological underpinnings of complex behavioral traits. Importantly, the applied analyses were carefully tailored to the specific research questions and integrated into robust pipelines for replication and validation of the initial results.