Rapid Screening of Anticoagulation Compounds for Biological Target-Associated Adverse Effects Using a Deep-Learning Framework in the Management of Atrial Fibrillation

The study introduces a deep-learning framework to screen anticoagulation compounds for adverse effects in atrial fibrillation management. Evaluated using SIDER and FAERS datasets, the model showed high precision and recall, particularly for enoxaparin and rivaroxaban. It identified stronger bleeding-related adverse effects for edoxaban compared to apixaban and enoxaparin. Sequoiaflavone's safety profile was similar to rivaroxaban. The framework aims to improve drug selection by predicting adverse effects early, potentially increasing clinical trial success rates. Future research should incorporate compound dose and patient-specific data to enhance the model's accuracy.