Robust Mendelian randomization methods incorporating weak and correlated instruments

Risk factors for common diseases and traits include epidemiological risk factors and genetic factors. Identifying causal epidemiological risk factors can accelerate innovations in disease prevention and intervention. MR methods are potentially powerful tools to investigate causal associations by using genetic variants as instrumental variables. However, these methods require strong assumptions that can lead to biased and inconsistent results. I develop robust and powerful MR methods to estimate and make inference on the causal effects between epidemiological risk factors and disease risks. The method requires weaker and more realistic assumptions and addresses several critical problems in available MR methods, including weak instrumental bias, pleiotropic effects, and linkage disequilibrium (LD) across variants. Through large-scale simulations under various genetic architectures, I have demonstrated the robustness and efficiency of this method.