Machine Learning Classification of TLR7 Ligands for Drug Discovery in Auto-immune Diseases

Toll-like receptor 7 (TLR7) plays a crucial role in antiviral immune responses and is strongly associated with autoimmune disorders such as systemic lupus erythematosus and rheumatoid arthritis. Due to the complexity of ligand–receptor recognition and the vast chemical search space, discovering selective TLR7 modulators remains highly challenging. In this study, a robust machine learning pipeline was developed using comprehensive molecular descriptors and Morgan fingerprints generated from a curated dataset of 5,115 compounds sourced from BindingDB and ChEMBL. Several classification algorithms were benchmarked using scaffold-based 5-fold cross-validation, ensuring reliable generalization to novel chemotypes. Among the evaluated models, the Random Forest classifier demonstrated superior predictive performance (accuracy = 0.934, precision =0.946, recall = 0.976, F1-score = 0.960), outperforming boosting-based and linear approaches. Independent external validation on 800 previously unseen molecules confirmed the model’s strong predictive power with an accuracy of 0.94, highlighting its robustness. Furthermore, feature importance analysis revealed key physicochemical properties and structural fingerprints governing TLR7 activation, offering interpretable insights for rational ligand design. Overall, the proposed ensemble learning framework provides an efficient and scalable strategy for prioritizing selective TLR7 modulators, facilitating high-throughput virtual screening and accelerating drug discovery toward immune-related therapeutic interventions.