Phenomewide scan of gene level associations
Principal Investigator: Dr Hae Kyung Im
Approved Research ID: 19526
Approval date: April 30th 2016
Genetic profile differences between individuals can lead to differences in molecular phenotypes such as gene expression levels. We hypothesize that the genetically modified differences in biomarkers can help us pinpoint genes involved in disease pathogenesis. In fact, we have developed a computational method called PrediXcan that detects genes associated with disease risk and related traits. Here we propose to apply this method across all phenotypes (baseline and derived from electronic medical records) available in the UK Biobank. These will be broadly shared through privacy-compliant resources such as web applications and databases. Our gene to phenotype associations will further our understanding of disease etiology. PrediXcan provides direction of the effects of gene alteration. This can allow us to prioritize drug targets since it is likely that down regulation of genes positively associated with disease risk lead to reduction of risk. We will use the genetic data on individuals to predict expression levels of genes (as well as other biomarkers) using prediction models generated by us in reference transcriptome datasets. These will be correlated with all the available and derived phenotypes. Analysis of phenotypes will be conducted to derive the best phenotypes (e.g. correction for environmental effects, longitudinal data may be summarized, electronic medical record-derived phenotypes). We will broadly share the results and build resources to further facilitate the downstream use of our results. Restrictions will be added as needed to protect the privacy of the participants. The largest sample size available will be used to maximize our power to detect genes associated with traits.