Genome-wide association studies and other forms of genome-phenome correlation require large numbers of individuals to achieve statistical validity and robustness, especially for complex traits. To support these kinds of studies across a broad range of human conditions, Vanderbilt has launched a biobank of DNA samples extracted from left-over blood after clinical testing for all patients who do not choose to opt-out of the collection. Samples are de-identified and linked to a 'synthetic derivative' database of de-identified data derived from electronic medical records. This presentation will outline the bioinformatics infrastructure and challenges associated with creation of a large scale de-identified genome-phenome correlation resource.