Fig. 8: Illustration for DeepSHAP implementations. | npj Precision Oncology

Fig. 8: Illustration for DeepSHAP implementations.

From: Autosurv: interpretable deep learning framework for cancer survival analysis incorporating clinical and multi-omics data

Fig. 8

DeepSHAP implementations to identify a latent features/clinical variables that contribute most to the difference in \({\rm{PI}}\) between high- and low-risk groups, and b pathways/genes/miRNAs that contribute most to the difference in latent-feature values (between high- and low-risk groups) for the most important latent features found in (a). SHAP Shapley Additive Explanations (SHAP) value (i.e., the contribution score in our setting).

Back to article page