Finding robust pathway markers for cancer classification

Advances in high-throughput measurement technologies have enabled the analysis of genome wide expression. One important problem in translational genomics is the identification of reliable and reproducible markers that can be used to effectively discriminate between different classes of a complex disease, such as cancer. The typical small sample setting makes the prediction of such markers very challenging. Recent studies have shown that pathway markers, which aggregate the gene activities in the same pathway, tend to be more robust than single gene markers and may improve the overall classification accuracy. To utilize pathway markers, we need a way to infer the activity level of a given pathway based on the expression of its member genes. In this work, we propose an improved pathway activity inference method that uses gene ranking to predict the pathway activity in a probabilistic manner. We show that the proposed method leads to better pathway markers with higher discriminative power and more consistent classification performance across different datasets.