Integer programming models and algorithms for molecular classification of cancer from microarray data

Novel, high-throughput technologies are challenging the core of algorithmic methods available in Computer Science. Microarray technologies give Life Sciencesresearchers the opportunity to simultaneously measure thousands of gene expression levels under different conditions or coming from different cell lines. With appropriate data mining models and algorithms, this would lead to a systematic exploration of molecular classification of cancer, just one among many other exciting applications. The aim of this paper is to present a unified mathematical formalization for different feature selection problems and investigate their performance in classification of cancer cell-lines. We also present some results using the NCI60 dataset.