Exploring the genetic association between multi-omic traits and psychiatric diagnosis, prognosis and treatment response

The biological mechanisms through which many behavioural traits express and develop are still unknown. This is particularly the case for psychiatric disorders, where diagnosis, prognosis and treatment decisions are not generally informed by molecular biomarkers but by behavioural questionnaires and tests. To better understand these disorders, it is essential to link them to their causal biological mechanisms. A specific study design is required to study the association between psychiatric disorders and biomolecular traits, where genetic information and tissue biomolecule concentrations are ideally measured in the same samples. These types of studies are very expensive, both in time and cost. However, advances in “omic” technologies (proteomics, metabolomics, transcriptomics) have enabled the generation of thousands of publicly available biomolecule blood concentration genome-wide association studies (GWAS). Recent work by Xu et al. (2023) highlights the role of these synthetic multi-omics GWAS datasets for the genetic prediction of disease. For this study, we have projected polygenic scores (PGS) for 17,227 blood concentration biomolecular traits into the iPSYCH data, a Danish genetics case-cohort study with over 90,000 psychiatric cases and a population cohort of 50,000 individuals. For a broad range of psychiatric diagnoses, this study aims to 1) perform a phenome-wide scan of associations to synthetic multi-omics datasets and 2) quantify the effect of these biomolecular traits on disease prediction. First, we projected PGS for the blood concentration of 17,227 biomolecular traits (proteins, metabolites and RNA) on the 134,677 individuals from the iPSYCH data. All PGS weights were developed by Xu et al. 2023 and are publicly available at omicspred.org. Briefly, the PGS were trained using Bayesian Ridge regression on the INTERVAL cohort, which comprised 50,000 genotyped healthy blood donors. Their blood samples have been processed using 5 platforms, including 867 metabolites, 2,692 proteins and 13,668 genes from Illumina RNA-seq. For the phenome-wide scan, models included birth year, sex and the 20 first principal components as covariates and FDR-corrected P Preliminary results indicate that 747 biomolecular traits were genetically associated with psychiatric diagnoses in iPSYCH. Of these, 594 were proteomic profiles, 97 were metabolic profiles and 56 were RNA expressions. For example, the largest effect sizes were for the repressed expression of TMEM161B in ADHD, the increased expression of PPP6C in mental disorders with psychoactive substance abuse and the repressed expression of PANK2 in anxiety-related disorders. Our preliminary results show that PGS for multi-omic traits are closely linked to psychiatric disorder diagnosis and have potential to discover new biological pathways of interest. A limitation of our study is absence of SNPs in iPSYCH that have INTERVAL PGS weights. We will explore ways to impute missing SNPs to finalise our analyses.