I am especially proud of a recent collaboration with Dr. Chad Shaw where we are developing applications of Causal Analysis to problems in clinical genetics and genomics. We envision a retranslation of classical genetics paradigms like ‘penetrance’ into the modern mathematical language of causality. Why is this necessary? A clinician who has ordered whole genome sequencing on a patient with suspected genetic disease is presented with >5 million variants in every case. Which one, if any, is responsible for the patient’s condition? We are on a very long road, but we hope that a solid mathematical foundation will clarify what we can say about molecular diagnosis and when we should proceed with caution.
The Causal Pivot (CP) is a structural causal model (SCM) for analyzing genetic heterogeneity in complex diseases. CP leverages one established causal factor to detect the contribution of a second suspected cause. Specifically, polygenic risk scores (PRS) serve as known causes, while rare variants (RV) or RV ensembles are evaluated as candidate causes. The CP incorporates outcome-induced association by conditioning on disease status. Understanding what happens when we select on disease status is particularly important in the molecular diagnosis scenario because the doctor always selects which patients are going to get genetic testing as well as the form of genetic testing. Dr. Shaw and his students derived a conditional maximum likelihood procedure for binary and quantitative traits and developed the Causal Pivot Likelihood Ratio Test (CP-LRT) to detect causal signals. Here is a link to our paper in the American Journal of Human Genetics:
belmontgenetics.org 