Skip to main content
. Author manuscript; available in PMC: 2022 Jan 25.
Published in final edited form as: J Chem Inf Model. 2020 Dec 21;61(1):46–66. doi: 10.1021/acs.jcim.0c00866

Figure 5.

Figure 5.

Comparing accuracy and interpretability among various versions of DeepAffinity+ (DeepAffinity with regularized and supervised attentions) and DeepRelations. “cstr” in legends indicates physical constraints imposed on attentions through regularization term R2(·), whereas “sup” indicates supervised attentions through regularization term R3(·). A horizontal dashed line indicates the performance of a random predictor.