🤖 AI Summary
This work addresses the inconsistency in predictions and explanations arising from multiple graph representations of the same molecule in molecular graph machine learning, which violates chemical identity. To resolve this, the authors propose InChIfied Invariants—strictly invariant features constructed at node, edge, and graph levels based on the International Chemical Identifier (InChI)—that inherently guarantee identical representations, predictions, and attributions for chemically equivalent graphs. Evaluation on the large-scale PubChem Substances dataset demonstrates that the method achieves consistent representations for 99.62% of chemically equivalent graph pairs, a dramatic improvement over the 0.35% consistency achieved by conventional Daylight invariants. Furthermore, it maintains competitive predictive performance on MoleculeNet benchmark tasks while significantly enhancing model interpretability and chemical plausibility.
📝 Abstract
Obtaining consistent explanations for machine learning on molecular graphs requires predictions and attributions to be aligned with chemical identity. However, chemically equivalent drawings of the same molecule can induce different molecular representations, leading to inconsistent predictions and explanations. Here, we introduce InChIfied Invariants, a class of node, edge, and graph features based on the International Chemical Identifier (InChI) and designed to be invariant under transformations that preserve chemical identity. Using one million molecular graphs from PubChem Substances, we show that InChIfied Invariants produce identical representations for chemically equivalent graphs in 99.62% of cases, whereas standard Daylight invariants do so in only 0.35% of cases. Across MoleculeNet tasks, InChIfied Invariants preserve predictive performance while significantly improving prediction consistency across alternative graph depictions of the same molecules. We further perform a quantitative attribution analysis and show that explanations produced with standard molecular featurization methods vary substantially across chemically equivalent graphs, while InChIfied Invariants enforce consistent attributions by construction. We release open-source software implementing InChIfied Invariants, which can be used as a drop-in replacement for standard molecular graph features.