The causal map of the human phenome (360G-Wellcome-208806_Z_17_Z)

This fellowship is focused on maximally exploiting GWAS summary data and 2SMR methodology. I will: 1. Create a computational framework that can construct and represent the causal map of the measured human phenome. Using existing (MR-Base) and upcoming (e.g. GoDMC, UK Biobank) data sources I will construct a graph of pairwise causal relationships between thousands of traits. Each causal relationship will be assessed using all available 2SMR and sensitivity methods to ensure transparency and highest standards for hypothesis-driven causal inference. 2. Exploit the graph to improve a) reliability of causal estimates by empirically assessing pleiotrpy; b) statistical power by exploiting outliers in the data to search for putative novel associations, and using graph traversal algorithms to find links for which direct associations are underpowered; c) interpretability by deconvolving the mediating relationships between arbitrary numbers of traits. 3. Use the causal graph as a tool for phenomic modelling. Example 1: linking molecular phenotypes to new or existing drug targets will enable the prediction of the efficacy and safety of developing interventions. Example 2: using MR against fitness related traits to obtain selection coefficients for each phenotype, I will construct models of human evolution, investigating the cost of complexity and evolutionary trajectories.