Can topology help neural nets predict how proteins behave?
Summarising 3D protein structures with topological fingerprints lets convolutional networks predict binding strength and mutation effects better than earlier methods, and multi-task learning helps on a small dataset.
Source
TopologyNet: Topology based deep convolutional and multi-task neural networks for biomolecular property predictions
Study at a glance
- Design
- Computational / modelling — Element-specific persistent homology features fed to convolutional and multi-task neural networks, evaluated on PDBBind, S2648/S350 and M223 benchmarks
- N
- Several benchmarks: 195-complex PDBBind 2007 core test set (1105 training), S2648 mutation set, and 223 membrane-protein mutations
- Population
- Protein–ligand complexes and protein mutation records
- Outcome
- Pearson correlation and RMSE for binding affinity and mutation-induced stability change
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
The authors converted 3D protein and protein–ligand structures into image-like 'barcodes' from element-specific persistent homology, which tracks connected components, rings and cavities across distance scales. These multichannel inputs were fed to convolutional networks to predict protein–ligand binding affinity and changes in protein stability after mutation. For membrane proteins, which have little data, a multi-task network was trained jointly with the larger globular-protein mutation dataset.
What they found
On the PDBBind 2007 benchmark the binding predictor outperformed all scoring functions it was compared with, and on the larger 2016 set it reached a median Pearson correlation of 0.81. Multi-task learning raised the membrane-protein correlation from 0.52 to 0.57. Features from higher topological dimensions added a little on top of pairwise-distance features, and averaging the neural network with a gradient-boosted tree model gave a small further gain.
The limits
What it doesn't show
The membrane-protein results remain weak, which the authors call not satisfactory, and the multi-task gain is modest on a small dataset of 223 mutations. Comparisons use numbers reported by other studies on fixed benchmark splits, so they do not show how the method handles proteins unlike those in the training data. The improvement from combining with tree models is marginal and concentrated on a few extreme cases.
Key terms
- Persistent homology
- A method that tracks when topological features like loops and cavities appear and disappear as a distance scale grows.
- Betti numbers
- Counts of topological features: components (Betti-0), rings (Betti-1) and cavities (Betti-2).
- Multi-task learning
- Training one network on related tasks together so a data-poor task can borrow representations from a data-rich one.
- Binding affinity
- How strongly a small molecule binds to a protein.
- Pearson correlation
- A measure from -1 to 1 of how linearly predictions track true values.
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
What do Betti-1 numbers count?
Common questions
Why not feed raw 3D coordinates to a CNN?
That is computationally costly and tangles geometric detail with biology; topological barcodes compress the structure into a structured, lower-dimensional input.
How did multi-task learning help membrane proteins?
Sharing convolution layers with the much larger globular-protein mutation task helped learn better features, raising correlation from 0.52 to 0.57.
Are neural networks strictly better than tree ensembles here?
Not strictly; they handled extreme values better, and averaging both gave the best results.
More on Protein structure prediction