Identification of Robust Pathway Markers for Cancer through Rank-Based Pathway Activity Inference

Figure 1

Probabilistic inference of rank-based pathway activity. For a given pathway, we first compute the ranking of the member genes for each individual sample in the dataset. Then we estimate the conditional probability mass function (PMF) of the gene ranking under each phenotype. Next, we transform the gene ranking into log-likelihood ratios (LLRs) based on the estimated PMFs and normalize the LLR matrix. Finally, the pathway activity level is inferred by aggregating the normalized LLRs of the member genes.