| The systematic documentation of human phenotypic information for genetic analyses is still in its infancy. Most studies currently phenotype their subjects with respect to only a handful of relevant traits, and the largest of these studies collect this information for a few hundred thousand individuals at most. Although restricted to traits concerning human health and disease, the clinical records of individual patients represent one of the best available resources for obtaining phenotypic information on a very large number of subjects. At the present time, however, such records are not typically linked with genotypic information, rendering most forms of genetic analyses intractable. In Chapter 1, I will briefly outline some of the current challenges associated with performing genetic analyses on clinical data. In Chapter 2, I will describe an approach that associates common, complex diseases with particular rare genetic variants by analyzing a collection of over 100 million patient records gathered from across the United States and Denmark. In Chapter 3, I will illustrate how this same approach can be used to detect widespread epistasis among particular genetic loci. In Chapter 4, I will describe the impact and extent of one of the many problems associated with the digital phenotyping of patients using clinical data---the widespread use of lexical and syntactic variation when referring to diseases in free text. Finally, Chapter 5 will provide a brief summary of these results and describe some of the future applications that clinical records may find within the field of human genetics. |
