FAFE: Immune Complex Modeling with Geodesic Distance Loss on Noisy Group Frames
This work addresses the problem of computational modeling of antibody-antigen complexes for researchers in structural biology and drug discovery, representing an incremental improvement over existing methods.
The paper tackles the challenge of accurately modeling antibody-antigen complexes by identifying a gradient vanishing issue in AlphaFold2's Frame Aligned Point Error (FAPE) loss function on high-rotational-error targets and proposes a novel geodesic loss called Frame Aligned Frame Error (FAFE) to optimize rotational and translational errors. By fine-tuning AlphaFold2 with this loss, they achieve a correct rate of 52.3% on an evaluation set and 43.8% on a low-homology subset, improving over AlphaFold2 by 182% and 100% respectively.
Despite the striking success of general protein folding models such as AlphaFold2(AF2, Jumper et al. (2021)), the accurate computational modeling of antibody-antigen complexes remains a challenging task. In this paper, we first analyze AF2's primary loss function, known as the Frame Aligned Point Error (FAPE), and raise a previously overlooked issue that FAPE tends to face gradient vanishing problem on high-rotational-error targets. To address this fundamental limitation, we propose a novel geodesic loss called Frame Aligned Frame Error (FAFE, denoted as F2E to distinguish from FAPE), which enables the model to better optimize both the rotational and translational errors between two frames. We then prove that F2E can be reformulated as a group-aware geodesic loss, which translates the optimization of the residue-to-residue error to optimizing group-to-group geodesic frame distance. By fine-tuning AF2 with our proposed new loss function, we attain a correct rate of 52.3\% (DockQ $>$ 0.23) on an evaluation set and 43.8\% correct rate on a subset with low homology, with substantial improvement over AF2 by 182\% and 100\% respectively.