Reliable OOD Virtual Screening with Extrapolatory Pseudo-Label Matching

1. 1Identify out-of-distribution regions in the data
2. 2Apply extrapolatory pseudo-label matching to improve model reliability
3. 3Evaluate model performance on held-out data
4. 4Refine model by incorporating additional data or adjusting hyperparameters

ML researchers and practitioners in drug discovery can benefit from this approach to improve the reliability of virtual screening, while data scientists and software engineers can implement and integrate this method into existing workflows

💡 Extrapolatory Pseudo-Label Matching can improve model reliability in out-of-distribution regions

💡 Improve reliability of OOD virtual screening with Extrapolatory Pseudo-Label Matching!