Principal Investigator: Professor Joanna Biernacka
Department: Mayo Clinic
Rochester, Minnesota, USATags: 55108, bipolar disorder, genetic heterogeneity, Machine Learning, polygenic risk score, psychiatric genetics, substance use disorder
Research has shown that psychiatric traits are influenced by many genetic variations with very small individual effects (i.e. these traits are highly polygenic), and statistical methods that account for the effects of many genetic variations (i.e. polygenic methods) have become widely used in psychiatric genetics research. Polygenic methods can help identify shared genetic risk factors across different disorders. For example, it has been shown that bipolar disorder shares genetic risk factors with schizophrenia. An understanding of the shared genetics across disorders can also help reveal genetic heterogeneity within disorders (e.g. schizophrenia genetic risk factors are shared especially with bipolar disorder with psychosis, and less so with bipolar disorder without psychosis). A better understanding of such shared genetics is improving our ability to identify the genetic contributors to these complex traits.
As a statistical genetics research group dedicated to understanding the genetic underpinnings of psychiatric traits and individual differences in response to their treatment, we use polygenic methods to study genetic heterogeneity that gives rise to disease subtypes, which can be defined by symptoms or comorbidities. To make further progress in characterizing the genetic heterogeneity of complex psychiatric diseases, we are extending currently used statistical methods to improve subtype prediction from genetic data. These methods need to be evaluated and applied in large datasets, and the UK Biobank will be an ideal dataset to advance this work. Specifically, we will:
1. Develop new analytical approaches, including extensions of polygenic risk score and machine learning methods, including new strategies for modeling gene-environment interactions, to improve genetic prediction of psychiatric traits. UK Biobank data will be used to evaluate the new methods.
2. Develop predictive models for psychiatric disorders, focusing on bipolar disorder and substance use disorder, and their intersection. We will consider predictive models that incorporate polygenic risk scores for multiple related traits and risk factor interactions, and will compare to prediction based on machine learning approaches. We will also use genetic data to model effects of possible intermediate risk factors such as gene expression and metabolite levels that may be altered in patients with psychiatric disorders, and incorporate these effects in psychiatric disorder prediction.
We expect our research to provide insights into genetic risk for psychiatric traits, which may ultimately improve clinical risk assessment and prediction of prognosis and treatment response. We also expect to develop statistical method that may prove useful in detecting genetic risk factors for other medical conditions.