(A) A principal component analysis was performed on 24 physicochemical features calculated on the pretraining set, the fine-tuned set and the generated molecular set. The first two principal components (PC1, PC2) were selected. (B) Scoring of molecules generated by reinforcement learning.
Mpro: Main protease; PLpro: Papain-like protease; QED: Quantitative estimate of drug-likeness; SA: Synthetic accessibility score.