Skip to main content
Fig. 5 | BMC Bioinformatics

Fig. 5

From: Network-based prediction approach for cancer-specific driver missense mutations using a graph neural network

Fig. 5

Interpretation of the results for Kim et al. dataset. This figure shows the prediction scores for “Cancer pathway + Molecular interaction” and “No graph,” and the graph feature importance rate for “Cancer pathway + Molecular interaction.” The circle plots are positive variants and square plots are negative variants in this benchmark dataset. The straight lines are thresholds for positive and negative for “Cancer pathway + Molecular interaction” and “No graph” ("Cancer pathway + Mo-lecular interaction" = 0.270, "No graph" = 0.218). The plots with red arrows were predicted negative incorrectly for “No graph” but positive correctly for “Cancer pathway + Molecular interaction”

Back to article page