Development of a knowledge-based-multi-scheme cancer microarray data analysis system

Comparing genes expressed in normal and diseased states assists the understanding of cancer pathophysiology, detection, prognosis, and therapeutic target study. Many existing expression analysis papers show that microarray data are usually case dependent, have small sample (patients) sizes, and have large gene dimensions. Thus, we have been developing a robust multi-parameter, multi-scheme knowledge-based optimization system that integrates the strengths of statistics, pattern-recognition, and support vector machines (SVM). The optimization logic identifies optimal cancer signature genes by utilizing different analysis models based on unsupervised and supervised clustering. Our system is being finalized by testing over public and in-house datasets with the intention of validation through clinical knowledge feedback.