Can a graph neural network judge predicted protein shapes?
Treating a protein as a graph of residues and contacts let a fairly simple neural network rank predicted protein structures about as well as the best existing single-model tools.
Source
GraphQA: protein model quality assessment using graph convolutional networks
Study at a glance
- Design
- Computational / modelling — Message-passing GCN trained on scored CASP 9-12 decoys, tested on CASP13 stage-2 decoys, with ablations over co-optimisation, depth/connectivity and features.
- N
- About 85k training/validation decoys from CASP 9-12; test set of roughly 14k decoys across 72 CASP13 targets; ablations averaged over 10 runs on CASP 11.
- Population
- Computationally generated protein structure models (decoys) from CASP editions
- Outcome
- Agreement between predicted and true quality scores (per-target Pearson R for global GDT_TS, Spearman correlation for per-residue LDDT/CAD, RMSE)
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The authors represented each predicted protein structure as a graph whose nodes are amino acids and whose edges link neighbouring residues in the chain or residues close in 3D space. A message-passing graph network was trained to predict both whole-model and per-residue quality scores on decoys from earlier CASP competitions, then tested on CASP13 targets. They ran ablations on network depth, contact cutoff, input features and whether local and global scores are learned together.
What they found
On CASP13, GraphQA ranked decoys within a target better than other single-model methods on the per-target correlation measure, and matched the best methods on per-residue scores. It clearly outperformed ProQ4, which uses the same input features but a 1D-CNN, and a variant using only amino-acid identity performed almost as well as the full model. Learning local and global scores together lowered global error, but did not help local error.
The limits
What it doesn't show
GraphQA was tested offline after the fact, whereas rival methods' numbers come from the blind CASP13 competition, so the comparison is not fully fair, as the authors themselves note. The graph uses only alpha-carbon positions, not full atomic detail. The refinement experiment is only a weak signal (a small average cosine similarity), so it does not show the model can actually improve structures.
Key terms
- Protein quality assessment (QA)
- Estimating how close a computationally predicted protein structure is to the true native structure.
- Decoy
- A candidate 3D model of a protein produced by a structure-prediction method.
- Graph convolutional network (GCN)
- A neural network that updates node and edge features by passing messages along the edges of a graph.
- Message passing
- Each layer lets nodes gather information from their neighbours, so stacking layers widens the receptive field.
- Ablation study
- Removing or changing one component at a time to measure how much it contributes to performance.
- GDT_TS
- The official CASP global score measuring how much of a model superimposes onto the native structure.
Flashcards
0 of 11 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
What do edges in the protein graph represent?
Common questions
Why represent a protein as a graph instead of a sequence or 3D grid?
A graph captures both chain order and spatial contacts, handles proteins of any size, and does not change when the protein is rotated or moved.
Is GraphQA proven better than CASP13 winners?
Not conclusively; it was evaluated after the competition, so a fair blind comparison would need a later CASP.
Does it need hand-engineered features?
They help, but a version using only amino-acid identity came close to the full model, showing the network learns useful features itself.
More on Graph neural networks