Background

Since the first microarray studies were published almost 15 years ago (DeRisi et al., 1997), advances in the ability to obtain diverse measurements from the genome have continued to occur at a rapid pace. Technological improvements have resulted in new methodologies for and increased efficiency of sequencing, phenotyping, and genotyping. These developments continue to increase the ease (and decrease the cost) of probing the genome and phenome of an individual. To date, however, little has been accomplished in the way of utilizing this rich source of data to make individualized decisions in the clinical setting. Although gene expression signatures have proven extremely useful in predicting outcomes in patients (e.g., breast cancer recurrence [Mook et al., 2007; Sparano and Paik, 2008] and colon cancer recurrence [Clark-Langone et al., 2010]), these approaches tend to categorize patients into a few groups and rely on a single source of genomic information. Personalized medicine, by definition, will require even more refined and specific categories, which will be more effective and informative if multiple sources of data are utilized.

Personalized genomic medicine seeks to fully characterize how genome and phenome heterogeneity relate to an outcome of clinical importance, such as response to treatment. Characterizing genome and phenome heterogeneity is of particular importance in cancer because the same disease can result from many different genomic events or abnormalities, and specific subgroups may have different treatment response. If we could catalog, for every patient, the specific genomic and downstream events that gave rise to cancer cells, this could be used to identify cancer subtypes.