Search papers, labs, and topics across Lattice.
This paper introduces MR-MoL, a multi-granular rationale-guided molecular LLM that enhances molecular property prediction by integrating GNN-derived attributions directly into the LLM's input. By scoring and ranking influential substructures, MR-MoL provides interpretable rationales that chemists can leverage, thereby bridging the gap between implicit molecular representations and explicit reasoning. The model outperforms generalist counterparts on eight MoleculeNet tasks and closely approaches the performance of specialized models, demonstrating its effectiveness in utilizing internal molecular cues for property prediction.
MR-MoL reveals how GNN-derived attributions can transform molecular property predictions by making the reasoning process transparent and interpretable.
Large language models (LLMs) are widely applied across chemical tasks, such as molecular property prediction, which underpins drug discovery. Molecular LLMs represent a molecule through several modalities, notably a 1D SMILES sequence or a 2D molecular graph. Both encode molecular information implicitly, so the contribution of individual substructures remains opaque. Retrieval and augmentation methods add context, but from external sources. However, the cues chemists reason over are the internal substructures that drive a property up or down. We propose MR-MoL, a multi-granular rationale-guided molecular LLM that supplies this evidence directly. A fine-tuned GNN scores each substructure through masking, and the most influential ones are serialized as a ranked, direction-tagged rationale that the LLM reads alongside the SMILES sequence and molecular graph. The rationale spans three levels of granularity: Murcko scaffolds with their side chains, BRICS fragments, and functional groups. This is, to our knowledge, the first method to expose GNN-derived attributions to an LLM as evidence for property prediction. On eight MoleculeNet tasks, MR-MoL achieves the best overall results among generalist models and narrows the gap to specialist models tuned for each task. Five diagnostics further confirm that the model reads the rationale rather than merely benefiting from its presence. Its direction, rank, and substructure each shape the prediction, and its attributions reproduce known structure-property relationships.
