Structure-Aware Fusion with Progressive Injection for Multimodal Molecular Representation Learning

TL;DR

MuMo introduces a structured multimodal fusion framework for molecular representation that combines 2D and 3D information with progressive injection, significantly improving robustness and performance across diverse molecular tasks.

Contribution

The paper presents MuMo, a novel fusion framework with a structured pipeline and progressive injection mechanism, enhancing stability and effectiveness in multimodal molecular modeling.

Findings

Achieves 2.7% average improvement over baselines on 29 tasks

Ranks first on 22 benchmark tasks, including a 27% boost on LD50

Demonstrates robustness to 3D conformer noise

Abstract

Multimodal molecular models often suffer from 3D conformer unreliability and modality collapse, limiting their robustness and generalization. We propose MuMo, a structured multimodal fusion framework that addresses these challenges in molecular representation through two key strategies. To reduce the instability of conformer-dependent fusion, we design a Structured Fusion Pipeline (SFP) that combines 2D topology and 3D geometry into a unified and stable structural prior. To mitigate modality collapse caused by naive fusion, we introduce a Progressive Injection (PI) mechanism that asymmetrically integrates this prior into the sequence stream, preserving modality-specific modeling while enabling cross-modal enrichment. Built on a state space backbone, MuMo supports long-range dependency modeling and robust information propagation. Across 29 benchmark tasks from Therapeutics Data Commons (TDC) and MoleculeNet, MuMo achieves an average improvement of 2.7% over the best-performing baseline on each task, ranking first on 22 of them, including a 27% improvement on the LD50 task. These results validate its robustness to 3D conformer noise and the effectiveness of multimodal fusion in molecular representation. The code is available at: github.com/selmiss/MuMo.