Dual-Attention Multimodal Framework for Molecular Property Prediction

Self-Supervised Learning for Molecular Property Prediction: Methods, Multimodal Insights, and Benchmark Comparisons.

Computer-aided drug discovery has substantially accelerated modern pharmaceutical research, where accurate molecular property prediction plays a central role in identifying promising therapeutic candidates. Self-supervised learning (SSL), which exploits large-scale unlabeled molecular data to learn transferable representations, has recently emerged as a powerful paradigm well-aligned with the data characteristics of cheminformatics. Integrating chemical domain knowledge further enhances the ability of SSL models to capture structural, physicochemical, and functional properties of molecules. In this review, we provide a systematic overview of recent advances in SSL-based molecular property prediction. We summarize representative methodological developments and analyze how multimodal molecular representation learning─by integrating sequence, graph, three-dimensional structure, and textual information─can improve the quality and expressiveness of molecular representations. We further examine the synergistic relationship between multimodal modeling and SSL, highlighting how complementary modalities can enhance representation learning in low-label settings. To demonstrate the practical benefits of multimodal molecular properties, we compare their performance with conventional SSL models on two downstream benchmark tasks with distinct prediction objectives. Finally, we discuss key open challenges, including the scarcity of high-quality 3D molecular data, modality imbalance across data sets, and the limited interpretability of learned representations. We conclude by outlining promising research directions toward more robust, generalizable, and biologically meaningful frameworks for molecular property prediction.

Attention based multi-scale fusion graphical framework for molecular docking of quinoline carboxamide analogues

Multimodal feature fusion for molecular property classification.

This study provides a large-scale empirical evaluation of multimodal feature fusion for molecular property classification by systematically integrating SMILES-based chemical language representations with fingerprint-based structural descriptors across 60 benchmark datasets.

MSMPP: Molecular Property Prediction by Integrating Multi-scale Multi-view information with pretrained 3D molecular large model representation.

Molecular property prediction is a cornerstone for accelerating drug discovery, providing a computational way to alleviate the low success rate, high cost, and long development cycle of conventional pharmaceutical research and development. However, existing computational methods have obvious limitations: most current deep learning approaches rely on a single molecular view, failing to fully capture the multi-dimensional features of molecular structures; they mainly focus on intra-molecular features while ignoring inter molecular information and cross-task correlations; and limited labeled data severely impairs their generalization to novel molecules. To address these issues, we propose MSMPP, a multi-scale, multi-view fusion framework for molecular property prediction that learns intra- and inter scale features simultaneously. For intra-scale feature learning, MSMPP integrates TxGemma-enhanced 1D sequence representations, Graph Transformer-derived 2D topological graph features, and Uni-Mol-derived 3D molecular conformational features. The Graph Transformer models long-range atomic dependencies, while the two pretrained models provide task-agnostic molecular prior knowledge from large-scale pretraining corpora, thereby improving generalization to novel molecules. For inter-scale feature learning, MSMPP constructs an inter-molecular graph (IMG) that explicitly models global pairwise interactions among chemically similar molecules and also extracts the cross-task features. Evaluations on eight MoleculeNet datasets show that MSMPP significantly outperforms state-of-the-art models, demonstrating its effectiveness in integrating multi-view intra-molecular features, inter-molecular features and cross-task information. Overall, MSMPP provides a competitive tool for molecular property prediction and supports the acceleration of drug discovery workflows.

Dual representation learning-based drug synergy prediction via sequence and molecular network reconstruction

Predicting drug synergy is crucial for optimizing drug combination therapies. However, it remains a challenging task to extract and integrate complex relational information from multiple data dimensions. This study proposes the DRL-DSP, a novel dual representation learning framework designed to enhance drug synergy prediction by integrating molecular-level features from SMILES sequences with graph-level relational information from reconstructed molecular networks. Our approach employs a SMILES-based data augmentation strategy, where randomized SMILES sequences are generated to enrich sequence representation diversity and enable a more comprehensive exploration of drug characteristics. Additionally, similarities between drug pairs are computed based on SMILES sequences to capture molecular relationships. A network is constructed for graph-level feature aggregation by integrating similarity-based edge weights into the adjacency matrix as an affinity matrix and incorporating feature matrices that reflect molecular properties. This combined representation improves the model's capability to capture molecular and relational information simultaneously. By utilizing encoding-decoding techniques for structural information of SMILES and convolution functions for molecular network representations, DRL-DSP integrates these complementary data resources to enhance drug synergy prediction. Furthermore, experiments under various conditions are conducted to verify the performance of DRL-DSP. Our approach addresses the limitations of single-modality methods and establishes a new paradigm for drug synergy prediction by integrating molecular and relational representations into a more effective and accurate framework.

HyperMolFusion: A Hypergraph-Enhanced Multi-Modal Fusion Framework for Accurate Molecular Property Prediction.

Molecular property prediction is a critical task in accelerating drug discovery. While deep learning has shown promise, prevailing single-modal methods struggle to integrate multi-source (e.g., atomic graph and molecular fingerprints), heterogeneous chemical knowledge, thereby failing to holistically represent molecular structures and capture the high-order synergistic interactions governing their functions. To address these challenges, we present HyperMolFusion, a hypergraph-enhanced multi-modal fusion model for molecular property prediction. Compared with traditional graphs limited to pairwise atomic bonds, HyperMolFusion models chemical motifs as hyperedges to explicitly capture high-order structural correlations and encode complex molecular interactions. The framework comprises three core representation learning modules: AtomConv for local atomic interaction learning via attention-enhanced message passing, HyperConv for motif-level high-order correlation extraction via hypergraph convolution with GRU gating, and a mixed molecular fingerprint module that adaptively integrates MACCS, PubChem, and Pharmacophore fingerprints. A chemically guided attention (CGA) mechanism then dynamically fuses these multi-level features into hierarchical molecular representations, alleviating over-smoothing and preserving structural information effectively. Evaluated on eight MoleculeNet benchmarks (covering regression and classification tasks), HyperMolFusion achieves promising performance. For regression, it achieves an RMSE of 0.611 in lipophilicity, 0.653 in ESOL, and 0.951 in FreeSolv. For classification, it achieves a ROC-AUC of 0.935 in ClinTox, 0.907 in BBBP, and 0.689 in SIDER. This work provides a systematic and effective solution for molecular property prediction: by holistically integrating atomic, motif, and global fingerprint information via hypergraph modeling, HyperMolFusion offers a more reliable computational tool to enhance the efficiency and accuracy of drug development pipelines.