Supplementary Material and software are available at the address http://biodcv.fbk.eu/listspy.html

No abstract

Summary In this paper we apply a predictive profiling method to genome copy number aberrations (CNA) in combination with gene expression and clinical data to identify molecular patterns of cancer pathophysiology. Predictive models and optimal feature lists for the platforms are developed by a complete validation SVM-based machine learning system. Ranked list of genome CNA sites (assessed by comparative genomic hybridization arrays – aCGH) and of differentially expressed genes (assessed by microarray profiling with Affy HG-U133A chips) are computed and combined on a breast cancer dataset for the discrimination of Luminal/ ER+ (Lum/ER+) and Basal-like/ER- classes. Different encodings are developed and applied to the CNA data, and predictive variable selection is discussed. We analyze the combination of profiling information between the platforms, also considering the pathophysiological data. A specific subset of patients is identified that has a different response to classification by chromosomal gains and losses and by differentially expressed genes, corroborating the idea that genomic CNA can represent an independent source for tumor classification.

In this paper, we present a method for predictive profiling of mass spectrometry data. The method integrates a spectra preprocessing pipeline with a complete validation setup aimed at identifying the discriminating peaks and at providing an unbiased estimate of the predictive classification error, based on SVM classifiers and on Entropy-based RFE procedure. A particular emphasis is placed upon avoiding selection bias effects throughout all the analysis steps, from preprocessing to peak importance ranking.

No abstract