StabiGAT

The Idea

This is a small side project that I developed in my free time and continue to work on whenever I have time.

I implemented a graph neural network (GNN) to predict the thermostability of $\beta$-sheet peptides by integrating structural, sequence-based, and simulation-derived features. For each peptide, a 3D structure was first predicted using ColabFold and subsequently subjected to 30 ns molecular dynamics simulations in explicit solvent using GROMACS with the AMBER99SB-ILDN force field. From these simulations, multiple global stability descriptors were extracted, including RMSD, radius of gyration, solvent-accessible surface area (SASA), hydrogen-bond counts, Lennard–Jones contacts, and potential energy, capturing dynamic aspects of peptide stability.

In parallel, residue-level representations were generated using ESM2 protein language model embeddings, providing contextual sequence information for each amino acid. Each peptide was then represented as a graph where nodes correspond to residues and edges were derived from an MD-based adjacency matrix reflecting persistent spatial contacts during the simulation. These local embeddings and global MD features were fused via a cross-attention mechanism and processed by multiple Graph Attention Network (GAT) layers to predict whether a peptide is thermally stable or unstable.

References

• Rodella, C., Lazaridi, S., & Lemmin, T. (2024) TemBERTure: advancing protein thermostability prediction with deep learning and attention mechanisms Bioinformatics Advances
  https://doi.org/10.1093/bioadv/vbae103

• Jung, F., Frey, K., Zimmer, D., & Mühlhaus, T. (2023) DeepSTABp: A Deep Learning Approach for the Prediction of Thermal Protein Stability International Journal of Molecular Sciences
  https://doi.org/10.3390/ijms24087444

• Li, B., Yang, Y. T., Capra, J. A., & Gerstein, M. B. (2020) Predicting changes in protein thermodynamic stability upon point mutation with deep 3D convolutional neural networks Computational Biology https://doi.org/10.1371/journal.pcbi.1008291

• Zhao, J., Yan, W., & Yang, Y. (2023) DeepTP: A Deep Learning Model for Thermophilic Protein Predictiona International Journal of Molecular Sciences https://doi.org/10.3390/ijms24032217

Third-party licenses

This project uses the packages:

• pytorch_geometric
• scikit-learn

MIT License

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.