An xQTL Map Integrates the Genetic Architecture of the Human Brain’s Transcriptome and Epigenome
This week we profile a recent publication in Nature Neuroscience from Dr. Bernard Ng
(above) and Dr. Sara Mostafavi (below) at the University of British Columbia.
Can you provide a brief overview of your lab’s current research focus?
At the Mostafavi Lab, we are developing novel statistical machine learning tools for analyzing multi-omic data. By applying these tools to large scale clinical datasets from our local and international collaborators, we are discovering new associations between biological layers, which helps further our understanding of psychiatric diseases.
What is the significance of the findings in this publication?
A key challenge to traditional genome wide association studies (GWAS) is that many identified genetic variants lie outside the protein coding regions, hence their functions are difficult to determine. To address this challenge, we performed quantitative trait locus (QTL) analyses on one of the largest multi-omic datasets for brain tissues, which comprises gene expression, DNA methylation, and histone modification data from ~500 older individuals. By associating all genetic variants along the genome to molecular features, a mapping of GWAS hits to their molecular functions is enabled. With this new resource, we showed that many xQTL SNPs influence multiple molecular features, and epigenetic mediation of eQTLs is found in some cases. We also demonstrated how this resource can be used to prioritize the cell type most affected by an xQTL, which is critical for designing follow-up molecular experiments where an in vitro or in vivo target cell type needs to be selected. Further, we re-analyzed 19 GWAS datasets with an xQTL-weighted approach and detected 20 new brain disease susceptibility loci. We are making this resource available for public use at the online portal, xQTL Serve: http://mostafavilab.
What are the next steps for this research?
The next step is to investigate whether additional xQTLs can be found by modeling the environment. For instance, individuals who smoke would have higher chance of getting lung cancer. Thus, we will examine whether certain genetic variants are associated with molecular features only when we consider their interactions with the environment. Rigorously establishing proxies for modeling various possible environmental factors will be a major challenge to this research.
This work was funded by:
We would like to acknowledge NSERC and BCCH for the NSERC DG grant and BCCH Establishment grant.