PROJECT SUMMARY Our long-term goal is to establish personalized structural biology – a precision medicine approach for inter- preting clinical sequencing data by jointly modeling all mutations in a patient’s proteome in the context of protein 3D structures, known human genetic variation, and other relevant data. In this project, we will develop the com- putational tools needed to integrate the wealth of available genetic variation data with cutting edge algorithms for efficiently modeling mutations to human protein structures and accurately quantifying their specific functional effects. This will provide a rich characterization of healthy and diseased proteomes and the means to generate actionable hypotheses about the effects of variants of unknown significance in individual patients. To demon- strate the power and relevance of this approach, we will apply it to facilitate variant interpretation in individuals in the Undiagnosed Diseases Network (UDN). We will then collaborate to validate our predictions. Our central hypothesis is that achieving the full promise of precision medicine requires the interpretation of a patient’s genetic variants in their 3D structural contexts and the integration of structural and clinical infor- mation. Patient genome interpretation is a major roadblock to fully realizing the transformative potential of per- sonalized medicine in the clinic. Current approaches for characterizing protein-coding variants of unknown sig- nificance have several shortcomings that limit their practical utility. First, they are not personalized; most are trained en masse on databases of known mutations across thousands of individuals. Thus, they are subject to ascertainment bias and ignore the background of other variants present in the individual. Second, most fail to provide specific biologically interpretable and thus therapeutically actionable predictions of a mutation’s effects beyond “benign” or “pathogenic”. Third, they are not stable and similar methods often disagree. Fourth, most are unable to interpret multi-base insertions and deletions. As a result and most importantly, current methods often give insufficient guidance to clinicians and fail to personalize next steps of treatment. Computational methods for modeling the effects of mutations on protein structures are now sufficiently fast and accurate to provide a solution to these challenges. Building on our expertise in analyzing the effects of mutations and modeling protein structures, the following aims establish a computational framework for interpre- tation of exonic variants that is personalized, clinically relevant, accurate, and applicable to all mutation types.