This past project is Roger’s doctoral research at University of Southern California, which focused on the development of machine learning and signal processing methods for analyzing gene expression patterns and copy number alterations in neuroblastoma tumors. In (Asgharzadeh, Pique-Regi et al., 2006, J. Natl. Cancer Inst.) we identified an expression signature based on 55 genes using a diagonally constrained linear discriminant analysis (DLDA). The classifier had an estimated accuracy rate of 85% significantly better than any other staging methods currently available. The methodology I developed received very positive comments from the journal editor (R. Simon, 2006 J. Natl. Cancer Inst): “The approach taken by the authors allowed them to avoid one of the major pitfalls of developmental studies [being published at that time], which is that they often provide highly biased estimates of accuracy…. Asgharzadeh et al. were careful to use cross-validation methods properly to avoid the large biases that can result from incomplete cross-validation.” In addition to changes in gene expression patterns we also explored copy number alterations. In (Pique-Regi et al., 2008 Bioinformatics) I developed a new computational method that exploits the signal processing properties of the copy number signal to: i) develop a maximally sparse representation for piece-wise constant vectors, and ii) use sparse Bayesian learning (SBL) to reconstruct the signal from noisy observations and recover the breakpoints. My approach, called genome alteration detection analysis (GADA), achieved one of the highest accuracies in breakpoint detection when compared to other popularly used copy number detection approaches while improving computational speed by several orders of magnitude.