The regulatory landscape of genetic variants associated with psychiatric disorders and neurodegenerative diseases
Date:
A. Amlie-Wolf , L. Qu, E.E. Mlynarski, C.D. Brown, G.D. Schellenberg, L.S. Wang
Genomewide association studies (GWAS) have identified thousands of mostly noncoding genetic variants associated with phenotypes, but due to linkage disequilibrium (LD), these variants tag many potentially causal variants. To identify the causal variants and tissuespecific regulatory mechanisms underlying GW AS signals, we propose the INFERNO (INFERring the molecular mechanisms of NOncoding genetic variants) pipeline. INFERNO builds LD blocks from GWAS variants using 1,000 Genomes Project data. Each linked variant is then overlapped with transcription factor binding sites (TFBSs) for 332 TFs and active enhancer annotations across 239 tissues and cell types from FANTOM5 and Roadmap Epigenomics. GWAS summary statistics are used in a Bayesian model to identify colocalized GWAS and eQTL signals across 44 GTEx tissues. Tissues and cell types are grouped into 32 tissue categories to allow data integration and provide unbiased identification of the affected tissue contexts. Background sampling is performed to quantify the enrichment of enhancer overlaps in each tissue category. To characterize the landscape of regulatory genetic variation across brainrelated traits, we applied INFERNO to variants associated with neurodegenerative diseases (Alzheimer’s disease (AD), amyotrophic lateral sclerosis (ALS), frontotemporal dementia, corticobasal degeneration (CBD)) and psychiatric disorders (schizophrenia (SCZ), bipolar disorder, attention deficit hyperactivity disorder, major depressive disorder, and autism). We identified significant enrichments of enhancer overlaps in the brain category in CBD, the brain and nervous categories in ALS, 18 categories including the blood and brain categories in AD, and 11 categories including blood, brain, and stem cell categories in SCZ. The blood category includes immunerelated cell types, providing evidence of an immune system role in AD and SCZ etiology . INFERNO identified 1,266 strongly colocalized GWASeQTL signals across these phenotypes, including 403 supported by variants overlapping concordant tissue enhancers with TFBS overlap and/or a high probability of underlying the colocalization signal. One AD signal has been validated by luciferase assays, with more experiments underway, and we continue to analyze the strongly colocalized signals. In summary, INFERNO provides a comprehensive and principled tool for integrating hundreds of diverse functional genomics datasets to characterize the regulatory effects of noncoding variants.