Identification of significant phenotypes related genes and biological pathways using a hybrid of support vector machines and smoothly clipped absolute deviation

A hybrid of support vector machines and the smoothly clipped absolute deviation (SVM-SCAD) is a penalized classifier that has been used by researchers especially in bioinformatics for defining significant genes in microarray analyzes. By combining support vector machines (SVM) with the smoothly clipped absolute deviation (SCAD), the SVM-SCAD has proven its ability to simultaneously select and classify genes that are related to phenotypes of interest. However, most researchers using genes equally a priory by neglecting the other existing biological data such as pathways data which may lead to a lack of interpretation for further biological explanations. Therefore, we propose the modified SVM-SCAD with group-specific parameter tunings in incorporating prior knowledge of pathways data. By incorporating this prior knowledge, not only are the significant genes defined, but pathways that are related to the phenotypes of interests are also detected, while group-specific parameter tunings allows flexibility in providing the power for selecting and discriminating informative genes within pathways. From the experiments using lung cancer microarray data, our proposed method shows that it is comparable to other penalized classifiers and other machine learning methods in defining significant genes and pathways that are related to phenotypes of interest. ICIC International