基于图神经网络、Deep & Cross网络与SMILES嵌入的多模态分子表示学习

查看论文PDF《基于图神经网络、Deep & Cross网络与SMILES嵌入的多模态分子表示学习》，作者Qiwei Han等3人

查看PDF HTML（实验性）

摘要：分子性质预测往往依赖孤立的数据模态，其中连续三维图神经网络（GNN）难以高效捕获长程拓扑依赖和精确的宏观启发式规则。在本工作中，我们提出了一种参数高效的三分支模块化融合神经网络，该网络综合了三种正交模态：三维空间几何（SchNet）、离散拓扑文法（通过ChemBERTa处理的SMILES）以及显式宏观物理化学描述符（Deep & Cross网络）。通过绕过标准标量读出并采用共享的晚期融合架构，该框架建立了一个数学上严格的多模态潜在空间，有效解决了局部消息传递中的算术和平滑过度限制。我们在QM9基准上评估了所提出的架构，目标为0 K下的原子化能（$U_0^{\mathrm{atom}}$）这一广泛热力学性质。通过系统的组合消融实验和潜在瓶颈优化（$d_e=64$），三模态框架达到了0.0207 eV的验证平均绝对误差（MAE）。该架构参数不足一百万，却果断超越了亚化学精度阈值，并且相较于严格控制的几何基线，误差大幅降低了20.6%。最终，我们的发现表明，整合正交的宏观和拓扑数据流提供了一种协同的$\mathcal{O}(1)$物理捷径。这种多模态对齐为粗暴的参数扩展提供了一种高效替代方案，为高通量虚拟筛选（HTVS）流程建立了鲁棒的代理模型。

投稿历史

来自：Qiwei Han [ 查看邮件 ] [v1] 2026年7月7日周二 01:47:44 UTC (98 KB)

[Submitted on 7 Jul 2026]

Title:Multimodal Molecular Representation Learning with Graph Neural Networks, Deep & Cross Networks, and SMILES Embeddings

View a PDF of the paper titled Multimodal Molecular Representation Learning with Graph Neural Networks, Deep & Cross Networks, and SMILES Embeddings, by Qiwei Han and 3 other authors

View PDF HTML (experimental)

Abstract:Molecular property prediction often relies on isolated data modalities, where continuous 3D graph neural networks (GNNs) struggle to efficiently capture long-range topological dependencies and exact macroscopic heuristics. In this work, we introduce a parameter-efficient Tri-Branch Modular Fusion Neural Network that synthesizes three orthogonal modalities: 3D spatial geometry (SchNet), discrete topological grammar (SMILES via ChemBERTa), and explicit macroscopic physicochemical descriptors (Deep & Cross Network). By bypassing standard scalar readouts and employing a shared late-fusion architecture, the framework establishes a mathematically rigorous multimodal latent space that effectively resolves the arithmetic and oversmoothing limitations of local message passing. We evaluate the proposed architecture on the QM9 benchmark, targeting the extensive thermodynamic property of atomization energy at 0 K ($U_0^{\mathrm{atom}}$). Through systematic combinatorial ablation and latent bottleneck optimization ($d_e=64$), the tri-modal framework achieves a validation Mean Absolute Error (MAE) of 0.0207 eV. Operating with fewer than one million parameters, this architecture decisively surpasses the sub-chemical accuracy threshold and yields a substantial 20.6% error reduction over a strictly controlled geometric baseline. Ultimately, our findings demonstrate that integrating orthogonal macroscopic and topological data streams provides a synergistic, $\mathcal{O}(1)$ physical shortcut. This multimodal alignment offers a highly efficient alternative to brute-force parameter scaling, establishing a robust surrogate model for high-throughput virtual screening (HTVS) pipelines.

Submission history

From: Qiwei Han [

view email

]

[v1]

Tue, 7 Jul 2026 01:47:44 UTC (98 KB)