AlphaSurf: On-the-Fly Surface Computations for Protein Representation Learning

Several protein surfaces have been proposed, notably for visualization purposes. Machine learning methods have incorporated these surfaces as protein representations, resulting in strong performances, at the cost of heavy computations. In this paper, we show that this burden can be avoided by introducing AlphaSurf, a coarse meshing method tailored for learning on proteins that relies on $\alpha$-complexes. Our method runs on-the-fly during training while maintaining performance. It also opens the door to data augmentation and learning on dynamics for surface-based methods.