Massively parallel sequencing technology has proven to enable the identification of driver genetic alterations in patients' tumors that may be suppressed by targeted therapies. Through retrospective analysis of clinical specimens, one can discover genomic biomarkers that predict outcomes and therapeutic response. Longitudinal profiling of multiple tumors in a single patient can reveal factors that influence tumor progression and drug resistance. Finally, prospective sequencing of patient specimens, when coupled with complementary radiology and histology based imaging, can enhance the clinical diagnosis and treatment of cancer patients. For increasingly lower costs, one can profile clinically relevant genes for mutations, copy number alterations, and structural rearrangements, with high detection sensitivity in low purity or multi-clonal tumor tissue. Advances in target capture, sample multiplexing, and profiling of formalin-fixed paraffin embedded (FFPE) specimens have further established the clinical utility of next generation sequencing. However, challenges remain in the application of these techniques to the analysis of clinical samples. In addition to the technical challenge of analyzing scant amounts of FFPE tissue, one must overcome the biological challenges of aneuploidy and heterogeneity inherent to the genetics of cancer. I will discuss different strategies for sequencing clinical samples, including different sequencing platforms, capture methods, and breadth of testing (i.e. targeted versus comprehensive approaches). I will describe examples in which our group has performed massively parallel sequencing on clinically annotated tumor specimens to identify genomic biomarkers of drug response and resistance. Finally, I will describe additional challenges in prospectively applying these techniques for clinical diagnosis involving bioinformatics, clinical interpretation, regulatory compliance, and ethics.