Predicting breast cancer using an expression values weighted clinical classifier

• DOI: 10.1186/s12859-014-0411-1
• Publication Date: 2014-12-30T13:07:08Z
• Authors: Minta Thomas; Kris De Brabanter; Johan AK Suykens; Bart De Moor
• Keywords: Cross-validation

Clinical data, such as patient history, laboratory analysis, ultrasound parameters-which are the basis of day-to-day clinical decision support-are often used to guide management cancer in presence microarray data. Several data fusion techniques available integrate genomics or proteomics but only a few studies have created single prediction model using both gene expression and These remain inconclusive regarding an obtained improvement performance. To improve management, these should be fully exploited. This requires efficient algorithms sets design final classifier. LS-SVM classifiers generalized eigenvalue/singular value decompositions successfully many bioinformatics applications for tasks. While bringing up benefits two techniques, we propose machine learning approach, weighted classifier sources: parameters. We compared evaluated proposed methods on five breast case studies. Compared individual sets, eigenvalue decomposition (GEVD) kernel GEVD, offers good performance, terms test area under ROC Curve (AUC), all Thus with set results significantly improved diagnosis, prognosis responses therapy. The has been shown promising mathematical framework non-linear classification problems.