FAFE: Immune Complex Modeling with Geodesic Distance Loss on Noisy Group Frames

Authors: Ruidong Wu, Ruihan Guo, Rui Wang, Shitong Luo, Yue Xu, Jiahan Li, Jianzhu Ma, Qiang Liu, Yunan Luo, Jian Peng

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental By fine-tuning AF2 with our proposed new loss function, we attain a correct rate of 52.3% (Dock Q > 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.
Researcher Affiliation Collaboration 1Helixon 2Tsinghua 3UTAustin 4Ga Tech.
Pseudocode Yes We provide an example implementation of F2E in Py Torch 2.0.1. (Appendix B.4 contains structured Python code snippets which serve as pseudocode for implementation).
Open Source Code Yes Code is available at https://github.com/mooninrain/FAFE.git.
Open Datasets Yes The fine-tuning dataset is collected from The Structural Antibody Database (Sab Dab) (Dunbar et al., 2014), which provides annotations on the original PDB database (Burley et al., 2017) raw structures.
Dataset Splits No The paper mentions a "training set" and "evaluation set" but does not specify the exact percentages or counts for a training/validation/test split, or reference predefined splits for this partitioning.
Hardware Specification No The paper mentions "computing resource constraints" but does not specify any particular hardware components like CPU or GPU models used for the experiments.
Software Dependencies Yes We provide an example implementation of F2E in Py Torch 2.0.1.
Experiment Setup Yes The batch size is set to 32, which we find is the minimal batch size to improve AF2 performance during fine-tuning. ... The crop length is set to 384 at max. The number of MSA cluster center is restricted to 128 for training efficiency.