Summary
Frame aligned point error (FAPE) is a loss function used to train protein structure prediction deep learning models such as AlphaFold2 and RosettaFold. It avoids the instabilities of training with RMSD. A modified version, frame-aligned frame error, was presented by (1) and fixes some problems regarding vanishing gradients when evaluating poses with large rotational deviations.
See also
1.
Wu R, Guo R, Wang R, Luo S, Xu Y, Li J, et al. FAFE: Immune Complex Modeling with Geodesic Distance Loss on Noisy Group Frames. In: International Conference on Machine Learning. PMLR; 2024. p. 53422–42. Available from: https://proceedings.mlr.press/v235/wu24g.html