De Novo Design of Nurr1 Agonists via Fragment-Augmented Generative Deep Learning in Low-Data Regime

Generative neural networks trained on SMILES can designinnovativebioactive molecules de novo. These so-called chemicallanguage models (CLMs) have typically been trained on tens of templatemolecules for fine-tuning. However, it is challenging to apply CLMto orphan targets with few known ligands. We have fine-tuned a CLMwith a single potent Nurr1 agonist as template in a fragment-augmentedfashion and obtained novel Nurr1 agonists using sampling frequencyfor design prioritization. Nanomolar potency and binding affinityof the top-ranking design and its structural novelty compared to availableNurr1 ligands highlight its value as an early chemical tool and asa lead for Nurr1 agonist development, as well as the applicabilityof CLM in very low-data scenarios.