A novel classification technique for cancer diagnostics based on microarray gene expression profiling (MGEP)

Pattern Recognition (PR) plays a vital role in the field of Bioinformatics. Various techniques of PR are used to analyze, segment and manipulate the high dimensional microarray gene expression data for classification. Microarray Gene Expression Profiling (MGEP) is an important domain of Bioinformatics that yields such high dimensional data used for various clinical applications such as cancer diagnostics and drug designing. In this study a novel scheme has been developed for the classification of unknown malignant tumors into known classes. The classification scheme includes the transformation of high dimensional microarray data on to the Mahalanobis space before classification, in order to compensate for the data spreads corresponding to each class and every gene. The efficiency of the proposed classification scheme has been proven on 10 publicly available cancer datasets containing both binary and multiclass data. To improve the performance of the classifier gene selection technique is applied on the datasets as a preprocessing/data extraction step.