Learning Inter-Graph Interactions Between Heterogeneous Substructures of Chemical Systems

Gyoung S. Na

cs.LGOct 20, 2026

Learning Inter-Graph Interactions Between Heterogeneous Substructures of Chemical Systems

Authors: Gyoung S. Na

Organizations: Korea Research Institute of Chemical Technology

Abstract

Complex chemical systems containing heterogeneous substructures are common in real-world applications. Various physical phenomena of the complex chemical systems are derived from the interactions between the heterogeneous substructures. However, existing graph representation learning methods for inter-graph interactions assumed graph-level interactions between homogeneous structures, such as organic molecules and inorganic crystalline materials. We propose a data descriptor of the complex chemical systems and a graph neural network for learning inter-graph interactions between organic and inorganic compounds. We applied the proposed method to predict the physical properties of hybrid solar cell materials containing heterogeneous substructures, which have received significant attention for sustainable energy resources. By learning heterogeneous inter-graph interactions, the proposed method achieved state-of-the-art accuracy in predicting band gaps of 1,682 hybrid solar cell materials.

Figures & tables

Figure 1: The overall structure of HOIPs. a: Two heterogeneous substructures of HOIP. b: Physical properties of HOIP from the physical interactions between inorganic frames and organic molecules.

Figure 2: The overall architecture of HeteroSIN for learning inter-graph interactions between two heterogeneous substructures of the input hybrid chemical system.

Dataset	GATv2	MPNN	CGCNN	UniMP	DimeNet++	M3GNet	HeteroSIN
CHOIP	0.846 ( $\pm$ 0.038)	0.931 ( $\pm$ 0.003)	0.928 ( $\pm$ 0.010)	0.895 ( $\pm$ 0.026)	0.902 ( $\pm$ 0.015)	0.948 ( $\pm$ 0.006)	0.965 ( $\pm$ 0.009)
THOIP	0.817 ( $\pm$ 0.037)	0.838 ( $\pm$ 0.024)	0.825 ( $\pm$ 0.043)	0.783 ( $\pm$ 0.087)	N/A	0.859 ( $\pm$ 0.036)	0.927 ( $\pm$ 0.046)

Table 1: Measured

R^{2}

-scores of the competitor GNNs and HeteroSIN in the problem of predicting the band gaps of HOIPs on the CHOIP and THOIP datasets. N/A means the negative

R^{2}

-score, which indicates the failure of the prediction.

Dataset	Zero Initialization	Gaussian Initialization	Initialization with I.F.	Initialization with q
CHOIP	0.913 ( $\pm$ 0.012)	0.935 ( $\pm$ 0.017)	0.915 ( $\pm$ 0.011)	0.965 ( $\pm$ 0.009)
THOIP	0.908 ( $\pm$ 0.035)	0.858 ( $\pm$ 0.051)	0.892 ( $\pm$ 0.048)	0.927 ( $\pm$ 0.046)

Table 2: Measured

R^{2}

-scores of HeteroSIN on the CHOIP and THOIP datasets for different initialization methods of the virtual node augmentation.

Graph Representation Learning

Graph representation learning develops methods for encoding nodes, edges, and whole graphs for prediction, comparison, and transfer.

See all 305 papers

ChemFlow:A Hierarchical Neural Network for Multiscale Representation Learning in Chemical Mixtures
Mar 3, 2026Jinming Fan, Chao Qian, Wilhelm T. S. Huck +2Molecular Property Prediction Molecular Representation Learning
College of Chemical and Biological Engineering, Zhejiang Provincial Key Laboratory of Advanced Chemical Engineering Manufacture Technology, Zhejiang University, 310027 Hangzhou (P. R. China) · Institute for Molecules and Materials, Radboud University, Nijmegen, The Netherlands · Institute of Zhejiang University – Quzhou, 324000 Quzhou (P.R. China)
MMGNN: Multi-level, multi-color graph neural networks for molecular property prediction
Jun 18, 2026Trung Nguyen, Duc Duy NguyenMessage Passing Neural Networks Molecular Property Prediction
The Bredesen Center, University of Tennessee, Knoxville, TN 37996, USA · Department of Mathematics, University of Tennessee, Knoxville, TN 37996, USA
Why Multi-Layer Message Passing Works: Completeness Theory for Graph Neural Network Interatomic Potentials
Sep 1, 2026Pingbing Ming, Han WangMachine Learning Interatomic Potentials GNN Expressivity
SKLMS, Institute of Computational Mathematics and Scientific/Engineering Computing, AMSS, Chinese Academy of Sciences, No. 55, East Road Zhong-Guan-Cun, Beijing 100190, China · National Key Laboratory of Computational Physics, Institute of Applied Physics and Computational Mathematics, Fenghao East Road 2, Beijing 100094, P.R. China, and HEDPS, CAPT, College of Engineering, Peking University, Beijing 100871, P.R. China

Mar 3, 2026J/K move · Enter open · S save

ChemFlow:A Hierarchical Neural Network for Multiscale Representation Learning in Chemical Mixtures

Jinming Fan, Chao Qian, Wilhelm T. S. Huck, William E. Robinson+1

College of Chemical and Biological Engineering, Zhejiang Provincial Key Laboratory of Advanced Chemical Engineering Manufacture Technology, Zhejiang University, 310027 Hangzhou (P. R. China) · Institute for Molecules and Materials, Radboud University, Nijmegen, The Netherlands · Institute of Zhejiang University – Quzhou, 324000 Quzhou (P.R. China)

Molecular Property Prediction Molecular Representation Learning Hierarchical Representation Learning Materials Science

Open paper AskPDF arXiv

Accurate prediction of the physicochemical properties of molecular mixtures using graph neural networks remains a significant challenge, as it requires simultaneous embedding of intramolecular interactions while accounting for mixture composition (i.e., concentrations and ratios). Existing approaches are ill-equipped to emulate realistic mixture environments, where densely coupled interactions propagate across hierarchical levels - from atoms and functional groups to entire molecules - and where cross-level information exchange is continuously modulated by composition. To bridge the gap between isolated molecules and realistic chemical environments, we present ChemFlow, a novel hierarchical framework that integrates atomic, functional group, and molecular-level features, facilitating information flow across these levels to predict the behavior of complex chemical mixtures. ChemFlow employs an atomic-level feature fusion module, Chem-embed, to generate context-aware atomic representations influenced by the mixture state and atomic characteristics. Next, bidirectional group-to-molecule and molecule-to-group attention mechanisms enable ChemFlow to capture functional group interactions both within and across molecules in the mixture. By dynamically adjusting representations based on concentration and composition, ChemFlow excels at predicting concentration-dependent properties and significantly outperforms state-of-the-art models in both concentration-sensitive and concentration-independent systems. Extensive experiments demonstrate ChemFlow's superior accuracy and efficiency in modeling complex chemical mixtures.

Abstract

Graph Representation Learning

Explore similar work

ChemFlow:A Hierarchical Neural Network for Multiscale Representation Learning in Chemical Mixtures

MMGNN: Multi-level, multi-color graph neural networks for molecular property prediction

Why Multi-Layer Message Passing Works: Completeness Theory for Graph Neural Network Interatomic Potentials