Screening of Inhibitors against Idiopathic Pulmonary Fibrosis: Few-shot Machine Learning and Molecule Docking based Drug Repurposing

Abstract

Introduction: Idiopathic pulmonary fibrosis is a chronic progressive disorder and is diagnosed as post-COVID fibrosis. Idiopathic pulmonary fibrosis has no effective treatment because of the low therapeutic effects and side effects of currently available drugs. Aim: The aim is to screen new inhibitors against idiopathic pulmonary fibrosis from traditional Chinese medicines. Methods: Few-shot-based machine learning and molecule docking were used to predict the potential activities of candidates and calculate the ligand-receptor interactions. In vitro A549 cell model was taken to verify the effects of the selected leads on idiopathic pulmonary fibrosis. Results: A logistic regression classifier model with an accuracy of 0.82 was built and, combined with molecule docking, used to predict the activities of candidates. 6 leads were finally screened out and 5 of them were in vitro experimentally verified as effective inhibitors against idiopathic pulmonary fibrosis. Conclusion: Herbacetin, morusin, swertiamarin, vicenin-2, and vitexin were active inhibitors against idiopathic pulmonary fibrosis. Swertiamarin exhibited the highest anti-idiopathic pulmonary fibrosis effect and should be further in vivo investigated for its activity.