Machine learning offers great promise for fast and accurate binding affinity predictions, however, current models lack robust evaluation and fail on tasks encountered in drug discovery. Here, the authors introduce an attention-based graph neural network model, AEV-PLIG, and demonstrate that augmenting the training data with semi-synthetic data significantly improves AEV-PLIG’s performance on benchmarks used for free energy perturbation calculations, narrowing the performance gap to simulation-based methods.
- Ísak Valsson
- Matthew T. Warren
- Philip C. Biggin