A primary objective in the study of human complex disease is to identify all risk factors that cause disease. Existing genetic association studies have succeeded in clarifying existing knowledge on which risk factors identified from correlational studies are truly causal for disease; for example, that elevated body-mass index (BMI) causally increases susceptibility to type-2 diabetes (T2D) and that lowering BMI can reduce T2D risk. But given the existing body of causal risk factors for a disease, can an individual’s overall risk be fully explained? If risk to a disease can be explained completely by this set of existing risk factors, this would be critical to demonstrate as it would shift research focus onto mechanistic studies designed to characterize those causal risk factors in greater detail. Alternatively, if there exists a portion of risk that can NOT be explained by known causal risk factors, this could imply potentially novel biology and/or mechanisms that remain to be discovered and characterized.
To address this question, our proposal seeks to statistically infer – from the set of observed genetic associations – if a hidden risk factor exists and, if so, place boundaries on the effect it could have on disease. To perform this inference, we are working to develop a new statistical method in the spirit of mediation analysis and its associated models to quantify the “unknown” contribution of a hidden risk factor to a disease’s genetic causal component. Quantifying the effect size of the hidden risk factors can potentially provide us insights on how far we are from fully understanding the disease and what the hidden risk factor could be. We are currently investigating the behaviors of the proposed methods via computer simulation studies, and our objective with this proposal for use of UK Biobank data is to apply the method to a collection of real data of diseases with known risk factors, including binary outcomes like type 2 diabetes, heart disease, as well as continuous traits including body mass index, height, and blood pressure.