Mutations in human proteins cause a wide spectrum of diseases and phenotypes.  Many of these mutations have been studied extensively and we now understand many of the basic biochemical functions these mutations disrupt, thereby leading to a clinical observation.  However, there are, according the Human Gene Mutation Database, now more than 38,000 protein mutations that are clinically associated with disease and, according to dbSNP, more than 40,000 known naturally occurring polymorphisms that result in an amino acid substitution.  These naturally occurring polymorphisms, for the most part, have not been annotated with any phenotype and their underlying effects are far from clear.  Understanding how a mutation causes biochemical changes that lead to a disease within the context of a specific patient is a formidable problem that will require years, perhaps decades, to unravel.  The goal of my bioinformatic research laboratory is to connect

models of amino acid function within a protein with mutations that cause disease to understand the molecular mechanisms of that disease.  One important effort in this area is developing binary classification tools that can discriminate from disease-causing (damaging) from non-disease-causing (neutral) mutations using protein, evolutionary and biochemical annotations.  Another important approach is building models of systems of how protein mutations affect biochemical systems.  In this

talk I will discuss our approaches in assembling and interpreting the biochemical effects of mutation data, as well as the challenge of noncoding variations and future approaches that will give us insight into the grand challenge of modeling genetic disease.