Studying genetics is an opportunity to understand biology with implications for treatment, diagnosis and prognosis. The degree to which a trait is genetic, or heritable, is a primary justification for genetic inquiry. Traits with high heritability provide opportunities for important biological insights and correspond to more value from the genetic information for potential clinical utility. Many factors affect heritability including the phenotype definition itself with simpler, more minimal approaches often yielding lower estimates of heritability. Electronic health records (EHR) provide a large feature space of clinical information in which to develop phenotypes and linked biobanks enable genetic studies at scale. In this talk, I will discuss statistical and machine learning approaches to leverage EHR data to derive more heritable phenotypes. These methods will be presented in the context of psychiatric phenotypes including suicide attempt and treatment resistant depression, among others.