Fig 1.
The reference gene is depicted by black circle. The initial static global PIN is projected onto normal and cancer samples based on gene expression, and each function (red and green) are diffused through each PIN. In this case, the reference gene is assigned green function in normal and red function in cancer, i.e., the gene gained red and lost the green function in cancer.
Table 1.
Functions ranked based on functional variability of genes.
Top 10 gained (green) and lost (red) functions are shown, along with Δf, Δf divided (normalized) by the number of genes annotated by the function, followed by the sample shuffling, and the log fold change, which is the log ratio of the average number of expressed genes annotated by f in cancer and normal samples.
Table 2.
Functions ranked based on expression variability of genes.
Top 10 (green) and bottom 10 (red) functions are shown based on log fold change of expression based activity (that is, number of annotated genes present in the corresponding projected PIN), along with Δf, Δf divided (normalized) by the total number of genes annotated by the function, followed by the sample shuffling results, and the log fold change.
Table 3.
Links between functional loss and mutation and deletion CNV.
The Fisher test contingency table showing the distribution of functions with elevated missense mutation frequency (columns 2 and 3) and deletion CNV rates (columns 4 and 5) between lost and gained functions. Mut(f) = 1 denotes significantly higher missense mutation frequency among the genes contributing to functional gain. CNV(f) = 1 has an analogous interpretation for deletion CNV.
Table 4.
Change in functional activity and association with patient survival.
Fisher test contingency table to test for association between functional loss/gain with associations with patient survival; β indicates the association of tumor-specific functional activity with survival risk.
Table 5.
Prediction accuracues using diffusion based fuctional profiles and annotation based functional profiles quantified by AUC-ROC.
The following table displays the AUC estimates of the 7 independent classifiers trained with two different feature sets (diffusion-based functional activity and annotation-based functional activity) for each clinical indicator.
Fig 2.
Clustering breast cancer samples based on their functional activity profile.
Kaplan-Meier survival curves of patients grouped in the 10 clusters show significant survival differences.
Fig 3.
Diffusion based functional heterogeneity across clinical subtypes.
The following figure displays the log ratio between the average numbers of genes assigned to each function by diffusion (represented by columns) across samples annotated with a subtype (represented by rows) versus the rest of the samples.
Table 6.
The table generated after performing diffusion based function assignment of a function to gene g in each tumor and normal sample.
Table 7.
The following table is generated to determine if elevated missense (respectively nonsense) mutation frequencies are enriched among functions with net gain (respectively net loss).