Fig. 7: Designed dual ligand candidates for AT1/sEH (10), FFAR1/sEH (11) and FFAR1/AT1 (12) from chemical language models.
From: Automated design of multi-target ligands by generative deep learning
a Synthesis of 10–12. b Target prediction (Z-scores) of designs obtained by beam search over the fine-tuning procedure using the Similarity Ensemble Approach (SEA)39 for the target of interest. Z-Scores are the mean ± standard deviation (SD). For each epoch beam search designs (width 50) were generated and only valid SMILES were analyzed. Source data are provided as a Source Data file.
