Multimodal molecular representation and property prediction based on structured complementary feature fusion
- 1College of Information Engineering, Hunan Open University , ,
Abstract
In this work, we propose Molrep-SCF, a multimodal self-supervised pretraining framework for molecular representation learning using SMILES sequences, 2D molecular graphs, and 3D geometric structures. The three modalities are encoded using a Transformer, a graph neural network, and an invariant 3D graph neural network, respectively. To enrich geometric supervision beyond a single conformation, Molrep-SCF incorporates three selected and complementary 3D conformational views. We further introduce a 2D-guided conformational consistency refinement mechanism that uses molecular topology to guide the fusion and refinement of their geometric representations. Instead of relying on a single static conformation, our method enforces structural consistency across conformers under the guidance of 2D topological priors. Furthermore, we design a structure-aware cross-modal learning strategy with atomic-level alignment via position-specific masking, enabling fine-grained correspondence across modalities. In addition, masked reconstruction across modalities enhances deep semantic coupling and facilitates information exchange between heterogeneous representations. Extensive experiments demonstrate that Molrep-SCF consistently outperforms existing multi-modal frameworks on molecular property prediction and other tasks, showing stronger generalization and representation quality.
Related
- Structured multimodal mobilization versus pilates-based motor imagery-focused pelvic floor exercises for primary dysmenorrhea: a randomized controlled trial
- Development of Improved Liver Steatosis Phantoms for Accurate Histopathologic Representation and Multimodal Imaging
- Stage-structured, distributional prediction of IVF outcomes with conditional updating
- TempoCross: instance-aware sparse representation for multimodal temporal fusion in 3D detection
- A Reproducible Decomposed Large Language Model Pipeline for Structured Abstraction of Multimodal Neuroimaging Reports: Development and Independent Validation in 1,669 Emergency Department Encounters for Dizziness and Vertigo
BibTeX
@article{ma2026multimodal,
title = {{Multimodal molecular representation and property prediction based on structured complementary feature fusion}},
author = {Meng Ma and Qingqing Cao and Fang Liu and Caihua Chen and Yan Liu},
journal = {Frontiers in pharmacology},
year = {2026},
month = sep,
doi = {10.3389/fphar.2026.1844064},
url = {https://europepmc.org/article/PMC/PMC13661354},
note = {CC BY},
}
RIS
TY - JOUR
TI - Multimodal molecular representation and property prediction based on structured complementary feature fusion
AU - Meng Ma
AU - Qingqing Cao
AU - Fang Liu
AU - Caihua Chen
AU - Yan Liu
PY - 2026
DA - 2026/09/30
JO - Frontiers in pharmacology
DO - 10.3389/fphar.2026.1844064
UR - https://europepmc.org/article/PMC/PMC13661354
AB - In this work, we propose Molrep-SCF, a multimodal self-supervised pretraining framework for molecular representation learning using SMILES sequences, 2D molecular graphs, and 3D geometric structures. The three modalities are encoded using a Transformer, a graph neural network, and an invariant 3D graph neural network, respectively. To enrich geometric supervision beyond a single conformation, Molrep-SCF incorporates three selected and complementary 3D conformational views. We further introduce a 2D-guided conformational consistency refinement mechanism that uses molecular topology to guide the fusion and refinement of their geometric representations. Instead of relying on a single static conformation, our method enforces structural consistency across conformers under the guidance of 2D topological priors. Furthermore, we design a structure-aware cross-modal learning strategy with atomic-level alignment via position-specific masking, enabling fine-grained correspondence across modalities. In addition, masked reconstruction across modalities enhances deep semantic coupling and facilitates information exchange between heterogeneous representations. Extensive experiments demonstrate that Molrep-SCF consistently outperforms existing multi-modal frameworks on molecular property prediction and other tasks, showing stronger generalization and representation quality.
SN - 1663-9812
ER -