Protein glycan docking accuracy improves with new dataset and model
ProtGlycanDock: Towards Accurate Protein-Glycan Docking with Tailored Dataset, Benchmark and Model
Machine Learning
Summary
Understanding how proteins and sugar molecules called glycans bind is important for biology but is hard to study because of a lack of good tools and data. The authors created a new dataset and benchmark specifically for testing protein-glycan docking models. They tested different methods and found AlphaFold 3 works best, then improved it by adding details about glycan shapes. Their new combined approach gave better results and helps others work on this problem.
What this means in practice
- •For drug designers: Predict accurate binding poses of protein-glycan complexes to aid drug development targeting glycan-related proteins.
- •For biotech developers: Integrate refined protein-glycan docking models into tools for designing glycan-based diagnostics or therapies.$Commercial implications: Enables development of commercial biotech software for protein-glycan interaction analysis, improving precision in therapeutic design.
Authors
Minghao Xu, Jiaze Song, Jinyu Wang, Wentao Zhang
Abstract
Performing accurate protein-glycan docking is critical to understanding protein-glycan interactions which govern many biological processes. However, this important task is hampered by the lack of tailored datasets, benchmarks and models for protein-glycan docking. To fill this blank, we first curate a standard benchmark ProtGlycanDock for training and evaluating protein-glycan docking models under both protein and glycan generalization settings. On this benchmark, we evaluate ten models in three categories, physics-based docking methods, machine-learning-based docking models and molecular structure prediction models. To obtain more precise binding poses, we further propose two finetuning techniques to inject the knowledge of glycan conformations into a general-purpose molecular structure prediction model. These two techniques respectively augment model inputs with glycan topological features and refine glycan binding pose with a glycan energy loss. Benchmark results show the superiority of AlphaFold 3 among existing methods, and the two proposed finetuning techniques can be well combined to achieve a state-of-the-art model for protein-glycan docking. Also, we perform stratified benchmark analysis and case studies to better understand the strengths and weaknesses of current models, providing insights for future model improvements. We maintain datasets at https://huggingface.co/datasets/ProtGlycanDock/ProtGlycanDock and source codes at https://github.com/kasawa1234/ProtGlycanDock