A simple, yet effective SVM based Protein expression pattern discovery for 2D-DIGE data: the Large-Scale Proteomics Ovarian Cancer analysis

FERNÁNDEZ, ELMER ANDRÉS; AMIUNE, HERNÁN; BALZARINI, MÓNICA

2009-01-01/2009-12-31

Metodológica

Ciencia y cultura-Ciencia y tecnologia

<!-- @page { margin: 0.79in } P { margin-bottom: 0.08in } --> Abstract—One of the main objective en biomarker discovery for cancer detection is to find the minimum amount of them that allow the best diagnosis/detection. One of the available high-throughput technologies to screen for protein biomarkers is 2D-DIGE. The usual protein expression detection schema is to rank the proteins according to some statistical test such as t-test, wilcoxon test, etc. Then the top (above a user threshold) proteins are selected as candidates. Here a SVM classification schema for protein detection is used after a t-test ranking. It is shown that high accurate classification rates can be achieved with different sets of ranked proteins, not necessary the top ones. It was devised a simple yet effective recursive SVM based method for protein selection. The proteins set achieve a near perfect classification over reduced set proteins compared to a previous published work.