Pathway Activity Inferences with Negatively Correlated Features for Pancreatic Cancer Classification

Pathway-based analysis has been extended to perform disease classification of expression profiles. In this study, gene sets of related pathways of pancreatic cancer from KEGG are used for microarray-based pancreatic cancer classification by using pathway activity inferences with negatively correlated features. Pearson´s correlation coefficient (PC) has been determined to be a suitable distance-based feature selection method with negatively correlated features. The results from classification performance when using gene sets in related pathways of pancreatic cancer suggest that not all related pathways are relevant to this disease, and the Jak-STAT signaling pathways is the most significant pathway. Pathway activity metrics of the proposed method with negatively correlated features can increase discriminative score and classification performance in many gene sets of related pathway of pancreatic cancer when compared to the original method of Lee etal. [7].