Logotyp: till Uppsala universitets webbplats

Depression is a very prevalent disorder affecting between 2 to 21% of the world population. This thesis extends the knowledge on the biological aspects of depression, aiming to identify and validate markers of genetic, epigenetic, and gene expression origin.

In Study I, the main focus was depression-related gene MAD1L1 that was previously linked to depression by SNPs and frequently mentioned as a stress-related marker. We identified that depression-related SNPs in MAD1L1 affect DNA methylation levels at cg02825527, cg18302629, and cg19624444 that were associated with depressive phenotypes in independent cohorts.

In Study II, we investigated whether GWAS catalog depression SNPs located in Olink-detectable genes could be replicated in a UKBiobank cohort and whether these associations are supported by DNA methylation and transcriptome. We validated eight depression SNPs and found very weak evidence that TNXB may be related to depression.

Study III was based on comparison of different depression -OMIC layers, including genetics, DNA methylation, and transcriptome. We explored how the identified genes from different -OMICs overlap, are functionally related and if they could show patterns in drugs and clinical trials. Only three genes were supported by evidence at all three -OMIC levels and included: FOXP1, VPS41, and AKTIP. Different -OMIC levels showed involvement of multiple systems in depression.

In Study IV, we used the Neuro Exploratory panel (Olink) to identify depression proteomic changes in blood. We took antidepressant intake into the account and validated associations in the independent datasets. We identified several proteins that showed nominally different levels between depression risk groups in the adolescent cohort. Validation of identified markers yielded that only PPP3R1 was also differentially expressed in prefrontal cortex and whole blood in the independent open-access cohorts with matching association directions.

In Study V, we used the entire blood DNA methylation as a depression marker. We investigated stability of DNA-methylation in eight independent datasets with meta-analysis and compared common machine learning and deep learning strategies for the depression detection purposes. We found 1987 CpG sites related to depression in both mega- and meta-analysis at the nominal level. Random forest classifiers achieved the best performance in identifying depression based on DNA methylation data in blood (AUC 0.73 and 0.76) in CV and hold-out tests respectively on the batch-level processed data.

Overall, the thesis supports multiple depression genetic, epigenetic, and expression markers. However, identified individual and systemic depression changes show high variability, which is in agreement with previous studies and observations.