Skip to main content
. 2022 Oct 4;62(20):4852–4862. doi: 10.1021/acs.jcim.2c00715

Figure 5.

Figure 5

Comparison of the model’s performance on the virtual screening benchmark when using an aggregate token embedding and the mean of all token embeddings for the molecular fingerprint. (A) shows the BEDROC20 score, and (B) shows the AUCROC score for each inference method.