Topological Feature Compression for Molecular Graph Neural Networks

2025/08/11 by Rahul Khorana, Khorana, Rahul
Computer Science · #Advanced Graph Neural Networks #Computational Drug Discovery Methods #FOS: Computer and information sciences #Graph Theory and Algorithms #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2508.07807

openalex publication_date 2025/08/11 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

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 \footnoteAll code and results can be found on Github https://github.com/rahulkhorana/TFC-PACT-Net.

Citations

• Analyzing Atomic Interactions in Molecules as Learned by Neural Networks
• Analyzing Atomic Interactions in Molecules as Learned by Neural Networks
• Polyatomic Complexes: A topologically-informed learning representation for atomistic systems
• Topological Neural Networks go Persistent, Equivariant, and Continuous
• Scientific discovery in the age of artificial intelligence
• Bi-level Contrastive Learning for Knowledge-Enhanced Molecule Representations
• How Powerful are K-hop Message Passing Graph Neural Networks
• Simplicial Attention Networks
• Multi-Scale Representation Learning on Proteins
• Pre-training Molecular Graph Representation with 3D Geometry
• Chemical-Reaction-Aware Molecule Representation Learning
• Non-Local Feature Aggregation on Graphs via Latent Fixed Data Structures
• Simple GNN Regularisation for 3D Molecular Property Prediction & Beyond
• Weisfeiler and Lehman Go Topological: Message Passing Simplicial\n Networks
• ForceNet: A Graph Neural Network for Large-Scale Quantum Calculations
• Performant implementation of the atomic cluster expansion (PACE): Application to copper and silicon
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• Hierarchical Message-Passing Graph Neural Networks
• Hierarchical Inter-Message Passing for Learning on Molecular Graphs
• Principal Neighbourhood Aggregation for Graph Nets
• Atomic Cluster Expansion: Completeness, Efficiency and Stability
• Optuna: A Next-generation Hyperparameter Optimization Framework
• How Powerful are Graph Neural Networks?
• SchNet – A deep learning architecture for molecules and materials
• Graph Attention Networks
• PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space
• Inductive Representation Learning on Large Graphs
• MoleculeNet: A Benchmark for Molecular Machine Learning
• Semi-Supervised Classification with Graph Convolutional Networks
• Adam: A Method for Stochastic Optimization
• Approximate Statistical Tests for Comparing Supervised Classification Learning Algorithms

Related

• Bi-level Contrastive Learning for Knowledge-Enhanced Molecule Representations
• Understanding the Capabilities of Molecular Graph Neural Networks in Materials Science Through Multimodal Learning and Physical Context Encoding
• Molecular Topological Profile (MOLTOP) -- Simple and Strong Baseline for Molecular Graph Classification
• GeomGCL: Geometric Graph Contrastive Learning for Molecular Property Prediction
• CTAGE: Curvature-Based Topology-Aware Graph Embedding for Learning Molecular Representations