XIMP: Cross Graph Inter-Message Passing for Molecular Property Prediction
This addresses the challenge of accurate drug discovery for researchers by improving generalization in low-data settings, though it is incremental as it builds on existing graph neural network methods with enhanced communication.
The paper tackled the problem of molecular property prediction in data-scarce regimes by introducing cross-graph inter-message passing (XIMP), which integrates multiple graph representations like molecular graphs and scaffold-aware junction trees, resulting in outperforming state-of-the-art baselines across ten diverse tasks.
Accurate molecular property prediction is central to drug discovery, yet graph neural networks often underperform in data-scarce regimes and fail to surpass traditional fingerprints. We introduce cross-graph inter-message passing (XIMP), which performs message passing both within and across multiple related graph representations. For small molecules, we combine the molecular graph with scaffold-aware junction trees and pharmacophore-encoding extended reduced graphs, integrating complementary abstractions. While prior work is either limited to a single abstraction or non-iterative communication across graphs, XIMP supports an arbitrary number of abstractions and both direct and indirect communication between them in each layer. Across ten diverse molecular property prediction tasks, XIMP outperforms state-of-the-art baselines in most cases, leveraging interpretable abstractions as an inductive bias that guides learning toward established chemical concepts, enhancing generalization in low-data settings.