Skip to main content
. 2023 Mar 23;63(7):1914–1924. doi: 10.1021/acs.jcim.2c01407

Figure 1.

Figure 1

Product, reactant, and reagent prediction models were trained with different combinations of input settings. Dashed boxes represent interchangeable configurations of data augmentation, tokenization and embedding strategies, molecule formats, and evaluation metrics.