Over the course of several years, we have carried out an analysis of a large-scale study on 168 blunt-force trauma patients over 28 days, measuring ~400 clinical variables and longitudinal gene expression with ~800 microarrays. The aim of this study is to improve our ability to diagnose and treat inflammatory complications in critically injured patients. In this talk, I will describe the challenges we have faced in analyzing these data, and some of the methodological developments we have made in addressing these challenges. I will specifically discuss methods for dealing with latent structure in gene expression data and approaches for modeling the data in such a way that confounding sources of variation are eliminated. The advances we have made from this study will likely be applicable to future large-scale clinical genomics studies, and they have directly motivated more general developments we have made in large-scale inference methods.