The incorporation of biological pathway information into microarray gene selection and classification has provided further biological information compared to traditional single gene analyses. One of the challenges in pathway-based microarray analysis is the quality of pathway data that leads to the presence of uninformative genes in the pathways. Another challenge is from the context free pathways information collection process where not all genes in the pathways are responsible for certain cellular process. Moreover, many algorithms in pathway-based microarray analysis neglect these limitations by including all the genes in a pathway and treated all the genes as significant. This can lead to a decrease in the statistical analysis performance. In order to overcome this challenge, a hybrid of support vector machines with smoothly clipped absolute deviation with groups-specific tuning parameter (gSVM-SCAD) method is proposed in this book. Experimental analyses using simulated data and two real canine and lung cancer microarray data have shown that the proposed method has the capability of identifying significant genes and pathways. The proposed method has also shown better results compared with other machine learning methods.