MOTIVATION: The drug-disease, gene-disease, and drug-gene relationships, as high-frequency edge types, describe complex biological processes within the biomedical knowledge graph. The structural patterns formed by these three edges are the graph motifs of (disease, drug, gene) triplets. Among them, the triangle is a steady and important motif structure in the network, and other various motifs different from the triangle also indicate rich semantic relationships. However, existing methods only focus on the triangle representation learning for classification, and fail to further discriminate various motifs of triplets. A comprehensive method is needed to predict the various motifs within triplets, which will uncover new pharmacological mechanisms and improve our understanding of disease-gene-drug interactions. Identifying complex motif structures within triplets can also help us to study the structural properties of triangles. RESULTS: We consider the seven typical motifs within the triplets and propose a novel graph contrastive learning-based method for triplet motif prediction (TriMoGCL). TriMoGCL utilizes a graph convolutional encoder to extract node features from the global network topology. Next, node pooling and edge pooling extract context information as the triplet features from global and local views. To avoid the redundant context information and motif imbalance problem caused by dense edges, we use node and class-prototype contrastive learning to denoise triplet features and enhance discrimination between motifs. The experiments on two different-scale knowledge graphs demonstrate the effectiveness and reliability of TriMoGCL in identifying various motif types. In addition, our model reveals new pharmacological mechanisms, providing a comprehensive analysis of triplet motifs. AVAILABILITY AND IMPLEMENTATION: Codes and datasets are available at https://github.com/zhanglabNKU/TriMoGCL and https://doi.org/10.5281/zenodo.14633572.

Predicting potential side effects of drug-drug interactions (DDIs), which is a major concern in clinical treatment, can increase therapeutic efficacy. In recent studies, how to use the multi-modal drug features is important for DDI prediction. Thus, it remains a challenge to explore an efficient computational method to achieve the feature fusion cross- and intra-modality. In this paper, we propose a dual-modality complex-valued fusion method (DMCF-DDI) for predicting the side effects of DDIs, using the form and properties of complex-vector to enhance the representations of DDIs. Firstly, DMCF-DDI applies two Graph Convolutional Network (GCN) encoders to learn molecular structure and topological features from fingerprint and knowledge graphs, respectively. Secondly, an asymmetric skip connection (ASC) uses distinct semantic-level features to construct the complex-valued drug pair representations (DPRs). Then, the complex-vector multiplication is used as a fusion operator to obtain the fine-grained DPRs. Finally, we calculate the prediction probability of DDIs by Hermitian inner product in the complex space. Compared with other methods, DMCF-DDI achieves superior performance in all situations using a fusion operator with the lowest parameter numbers. For the case study, we select six diseases and common side effects in clinical treatment to verify identification ability of our model. We also prove the advantage of ASC and complex-valued fusion can achieve to align the cross-modal fused positive DPRs through a comprehensive analysis on the phase-modulus distribution histogram of DPRs. In the end, we explain the reason for alignment based on the similarity of features and node neighbors.

In biomedical networks, molecular associations are important to understand biological processes and functions. Many computational methods, such as link prediction methods based on graph neural networks (GNNs), have been successfully applied in discovering molecular relationships with biological significance. However, it remains a challenge to explore a method that relies on representation learning of links for accurately predicting molecular associations. In this paper, we present a novel GNN based on link representation (LR-GNN) to identify potential molecular associations. LR-GNN applies a graph convolutional network (GCN)-encoder to obtain node embedding. To represent associations between molecules, we design a propagation rule that captures the node embedding of each GCN-encoder layer to construct the LR. Furthermore, the LRs of all layers are fused in output by a designed layer-wise fusing rule, which enables LR-GNN to output more accurate results. Experiments on four biomedical network data, including lncRNA-disease association, miRNA-disease association, protein-protein interaction and drug-drug interaction, show that LR-GNN outperforms state-of-the-art methods and achieves robust performance. Case studies are also presented on two datasets to verify the ability to predict unknown associations. Finally, we validate the effectiveness of the LR by visualization.