Does building 3D symmetry into a GNN improve protein site prediction?
A graph neural network whose outputs transform correctly under rotations, reflections and translations predicted protein interaction sites better than earlier methods, even from AlphaFold2-predicted structures.
Source
E(3) equivariant graph neural networks for robust and accurate protein-protein interaction site prediction
Study at a glance
- Design
- Computational / modelling — EquiPPIS (E(3)-equivariant GNN) trained on 335 proteins, tested on 60, compared with nine prior predictors plus invariant/no-attention/GCN/GAT ablations; also tested on unbound and AlphaFold2 structures
- N
- N=60 · 60 test proteins (Test_60); 335 training proteins; 31-protein unbound test subset; 42-protein validation set
- Population
- Protein chains from three public PPI benchmark datasets
- Outcome
- Per-residue interaction-site classification: ROC-AUC, PR-AUC, F1, MCC
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The authors turned each protein into a graph of residues linked when their alpha-carbons lie within 14 angstroms, and fed coordinates plus sequence and structure features into an E(3)-equivariant graph neural network (EquiPPIS) that labels each residue as an interaction site or not. They trained on 335 proteins and tested on 60, against nine existing predictors. To test why it worked they trained versions with the coordinate updates switched off (making it merely invariant), with attention removed, and ordinary GCN and GAT networks, all with the same features.
What they found
EquiPPIS was the only method to exceed a ROC-AUC of 0.8 on the test set, beating all sequence- and structure-based competitors. The invariant version and plain GCN/GAT networks did worse, pointing to equivariance as the main source of the gain, with attention adding a smaller boost. Using AlphaFold2-predicted structures, EquiPPIS reached a PR-AUC of 0.451, higher than the best competitor achieved even with experimental structures (0.429), and it lost less accuracy on unbound structures.
The limits
What it doesn't show
The test set is small (60 proteins, 31 in the unbound subset), so the gaps between methods are estimated from few examples, although the authors ran resampling significance tests. The method predicts sites for a single protein without knowing its partner, so it cannot say which partner binds where. The authors note the network is not interpretable, and they did not test adding multiple sequence alignment information.
Key terms
- Equivariance
- A property where transforming the input (for example rotating it) transforms the output in the matching way, rather than changing the prediction arbitrarily.
- E(3) group
- The set of rotations, reflections and translations in three-dimensional space.
- Invariant network
- A model whose output does not change at all under those transformations; here created by switching off coordinate updates.
- Graph neural network
- A network that passes messages between connected nodes of a graph to learn node or graph-level representations.
- PR-AUC
- Area under the precision-recall curve, a metric suited to imbalanced problems where positives are rare.
Flashcards
0 of 10 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Which transformations does an E(3)-equivariant network respect?
Common questions
Why does equivariance matter for protein structures?
A protein's function doesn't depend on how it is oriented in space, so a model that respects rotations and reflections by design doesn't have to learn that from data or augmentation.
How did the authors show equivariance was responsible for the improvement?
They retrained the same model with coordinate updates turned off, keeping features and hyperparameters fixed, and that invariant version performed worse.
Why test on AlphaFold2 models?
Most proteins lack experimental structures, so working well on predicted structures means the method can be applied at proteome scale.
More on Graph neural networks