Abstract:Abstract Herbs applicability in disease treatment has been verified through experiences over thousands of years. The understanding of herb–disease associations (HDAs) is yet far from complete due to the complicated mechanism inherent in multi-target and multi-component (MTMC) botanical therapeutics. Most of the existing prediction models fail to incorporate the MTMC mechanism. To overcome this problem, we propose a novel dual-channel hypergraph convolutional network, namely HGHDA, for HDA prediction. Technically, HGHDA first adopts an autoencoder to project components and target protein onto a low-dimensional latent space so as to obtain their embeddings by preserving similarity characteristics in their original feature spaces. To model the high-order relations between herbs and their components, we design a channel in HGHDA to encode a hypergraph that describes the high-order patterns of herb-component relations via hypergraph convolution. The other channel in HGHDA is also established in the same way to model the high-order relations between diseases and target proteins. The embeddings of drugs and diseases are then aggregated through our dual-channel network to obtain the prediction results with a scoring function. To evaluate the performance of HGHDA, a series of extensive experiments have been conducted on two benchmark datasets, and the results demonstrate the superiority of HGHDA over the state-of-the-art algorithms proposed for HDA prediction. Besides, our case study on Chuan Xiong and Astragalus membranaceus is a strong indicator to verify the effectiveness of HGHDA, as seven and eight out of the top 10 diseases predicted by HGHDA for Chuan-Xiong and Astragalus-membranaceus, respectively, have been reported in literature.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to predict the associations between Chinese herbs and diseases (Herb - Disease Associations, HDAs) more accurately. Due to the characteristics of multi - target and multi - component (Multi - Target Multi - Component, MTMC) of Chinese herbs, most of the existing prediction models fail to effectively incorporate this complex mechanism, resulting in poor prediction effects. To solve this problem, the author proposes a new dual - channel hypergraph convolutional network (Dual - channel Hypergraph Convolutional Network), namely HGHDA (Hypergraph - based Herb - Disease Association prediction), to better capture the complex MTMC relationships between Chinese herbs and diseases, thereby improving the prediction accuracy. ### Specific background of the problem 1. **Complex relationships between Chinese herbs and diseases**: - Chinese herbs have been proven effective in treating various diseases through thousands of years of practice. - However, the multi - target and multi - component characteristics of Chinese herbs make their mechanisms very complex and difficult to fully understand. - Most of the existing prediction models do not fully consider this complexity, resulting in inaccurate prediction results. 2. **Application of network pharmacology**: - The concept of network pharmacology has received extensive attention in recent years. It explores the action mechanisms of Chinese herbs in treating different diseases by constructing heterogeneous information networks. - This method provides a new perspective for predicting the associations between Chinese herbs and diseases. 3. **Limitations of existing methods**: - Existing methods mainly focus on predicting the associations between components - target proteins or Chinese herbs - target proteins, and pay less attention to the direct associations between Chinese herbs and diseases. - These methods often fail to effectively utilize the MTMC characteristics of Chinese herbs. ### Proposed solutions To overcome the above problems, the author proposes the HGHDA model, and the main features of this model are as follows: 1. **Dual - channel hypergraph convolutional network**: - HGHDA adopts a dual - channel structure to model the high - order relationships between Chinese herbs and their components and the high - order relationships between diseases and their target proteins respectively. - Each channel captures these high - order patterns through hypergraph convolution operations, thereby better reflecting the MTMC characteristics of Chinese herbs. 2. **Auto - encoder for dimension reduction**: - Use an auto - encoder to project the similarity matrices of components and target proteins into a low - dimensional latent space to generate embedding representations and retain the similarity features in the original feature space. 3. **Scoring function for prediction**: - Finally, a scoring function is used to predict the potential associations between Chinese herbs and diseases. ### Experimental verification The author conducted extensive experiments on two benchmark datasets, and the results show that HGHDA is superior to many existing advanced algorithms in prediction performance. In addition, through case studies of Ligusticum chuanxiong and Astragalus membranaceus, the effectiveness of HGHDA is further verified. In conclusion, this paper aims to solve the problem that existing prediction methods cannot effectively capture the complex mechanisms of Chinese herbs by introducing the HGHDA model, thereby improving the accuracy of predicting the associations between Chinese herbs and diseases.

Dual-channel hypergraph convolutional network for predicting herb–disease associations

ReHoGCNES-MDA: prediction of miRNA-disease associations using homogenous graph convolutional networks based on regular graph with random edge sampler

Predicting Potential Drug-Disease Associations Based on Hypergraph Learning with Subgraph Matching

Hierarchical and dynamic graph attention network for drug-disease association prediction

HeteroDualNet: A Dual Convolutional Neural Network With Heterogeneous Layers for Drug-Disease Association Prediction via Chou’s Five-Step Rule

Dynamic category-sensitive hypergraph inferring and homo-heterogeneous neighbor feature learning for drug-related microbe prediction

HTINet2: herb-target prediction via knowledge graph embedding and residual-like graph neural network

Prediction of Drug-Disease Associations Based on Multi-Kernel Deep Learning Method in Heterogeneous Graph Embedding

Predicting herb-disease associations using network-based measures in human protein interactome

MCHAN: Prediction of Human Microbe-drug Associations Based on Multiview Contrastive Hypergraph Attention Network

Predicting drug-disease associations with heterogeneous network embedding.

Predicting miRNA–Disease Associations by Combining Graph and Hypergraph Convolutional Network

MHCLMDA: multihypergraph contrastive learning for miRNA–disease association prediction

HGSMDA: miRNA-Disease Association Prediction Based on HyperGCN and Sørensen-Dice Loss

Hierarchical Negative Sampling Based Graph Contrastive Learning Approach for Drug-Disease Association Prediction

Homogeneous Symptom Graph Attentive Reasoning Network for Herb Recommendation

MGATMDA: Predicting Microbe-Disease Associations via Multi-Component Graph Attention Network

Gndd: A Graph Neural Network-Based Method For Drug-Disease Association Prediction

Drug Repositioning Based on the Enhanced Message Passing and Hypergraph Convolutional Networks

Predicting potential microbe-disease associations based on dual branch graph convolutional network

HGCLMDA: Predicting mRNA-Drug Sensitivity Associations via Hypergraph Contrastive Learning