Topological Feature Compression for Molecular Graph Neural Networks
This work addresses a major problem in cheminformatics and bioinformatics for researchers and practitioners by providing a parameter-efficient model that improves molecular prediction tasks, though it appears incremental as it builds on existing GNN methods.
The paper tackles the challenge of balancing predictive accuracy, interpretability, and computational efficiency in molecular representation learning by introducing a Graph Neural Network architecture that combines compressed higher-order topological signals with standard molecular features, achieving best-performing results in accuracy and robustness across almost all benchmarks.
Recent advances in molecular representation learning have produced highly effective encodings of molecules for numerous cheminformatics and bioinformatics tasks. However, extracting general chemical insight while balancing predictive accuracy, interpretability, and computational efficiency remains a major challenge. In this work, we introduce a novel Graph Neural Network (GNN) architecture that combines compressed higher-order topological signals with standard molecular features. Our approach captures global geometric information while preserving computational tractability and human-interpretable structure. We evaluate our model across a range of benchmarks, from small-molecule datasets to complex material datasets, and demonstrate superior performance using a parameter-efficient architecture. We achieve the best performing results in both accuracy and robustness across almost all benchmarks. We open source all code \footnote{All code and results can be found on Github https://github.com/rahulkhorana/TFC-PACT-Net}.