Fig 1.
A method to analyze the coupling of pathway activity on patient survival.
We used singular value decomposition to give, for each sample and pathway, an individualized pathway readout. We then combine those readouts with the survival information to perform survival analysis on a pathway level.
Fig 2.
Heatmap of the coefficients of the Pearson correlation matrix between the variables survival time, the eigengene of the gene expression values, and the gene expression values within the pathway ‘Metabolism of folate and pterines” for samples from diseased patients.
We sorted the genes in the expression matrix by their contribution to the eigensample and encompassed them with a dotted line. The patients’ survival time both correlates and anti-correlates with the different genes. However, the eigengene captures the covariational trend within the expression data and correlates well with the patients’ survival time.
Fig 3.
Comparison of how well eigengenes predict survival.
A histogram of the concordance indices was calculated with 5-fold cross-validation of Cox regressions based on eigengenes compared to (A) individual transcripts and (B) ssGSEA enrichment scores.
Fig 4.
Sunburst diagrams offer an exploratory visualization of the influence of the pathways in the Reactome hierarchy on the survival of patients.
We use an interactive sunburst chart to visualize and explore the result. This allows us to navigate the pathway hierarchy and see the relationships in a more natural way than e.g. in a static table. The sunburst (A) without and (B) with the pathway “Diseases of mitotic cell cycle” regressed out. Interactive versions of the same figures are available from https://statisticalbiotechnology.github.io/metabric-pathway-survival.