The relationship between microbes and human health is crucial for understanding pathogenesis and advancing clinical drug development. Investigating the microbe and disease associations(MDA) is essential in this regard. Computational methods offer a cost-effective and efficient approach to detect interactions among biological entities. We propose GCLMDA, a model that utilizes Graph Contrastive Learning for MDA prediction. Firstly, we gather data from diverse and heterogeneous sources to construct similarity networks for microbes and diseases respectively. Moreover, the graph convolution networks help us to learn multi-view node embeddings which are further refined through contrastive learning. Finally, the interacted features of nodes are input into a binary classifier to predict potential MDAs. Through extensive experiments on two public datasets, the proposed model consistently outperforms baseline methods, achieving the highest AUC and AUPR scores of 0.9569 and 0.9595, respectively. The case study on Type II Diabetes provides further evidence of the model effectiveness.

Fact-checking fake news detection involves using verified accurate factual information in news reports as "evidence" to validate objective statement "claim". Existing works primarily focus on identifying critical elements within the evidence that support or refute specific claims by assessing the congruence or divergence between the claim and the associated evidence. These methods can broadly be divided into text-based and graph-based-the former centers on understanding the nuances of unstructured text to extract semantic word-level information. At the same time, the latter is proficient at analyzing the node-level structure of graphs it creates from the text to reveal topological insights. Each type provides a distinct view on identifying critical elements for fact-checking. To enhance the complementary nature of the two perspectives, this paper proposes an end-to-end framework for fact-checking fake news detection entitled Multi-view Counterfactual Contrastive Learning (MCCL). The framework incorporates a counterfactual technique to refine the fused features from both the "entity-view" of textual content and the "centrality-view" of the graph structure. Additionally, it employs contrastive learning to sharpen the distinctions among multi-view features, which facilitates the exact identification of critical elements in the evidence related to their respective claims. Experimental results on real datasets demonstrate that the proposed MCCL outperforms state-of-the-art methods.

Drug repositioning is a promising strategy for treating diseases by repurposing safe drugs, which reduces development costs and timelines. Drug-disease association (DDA) plays a crucial role in this strategy. Although existing DDA models have made significant progress, there is a need for improvement in integrating diverse information and creating better training data. In this paper, we propose a novel method, termed DNSDDA that utilize the contrast learning and denoised negative sampling strategy for drug-disease association prediction. Firstly, DNSDDA constructs semantic networks from multiple data sources and meta-path networks from topological information for drugs and diseases. Then DNSDDA enhances topological embeddings of drugs and diseases by infusing semantic information through contrastive learning in the node representation of the meta-path network. After that, we employed denoised negative sampling to enhance the training set quality. Lastly, DNSDDA trains a binary classifier using learned embeddings of drugs and diseases for predicting potential associations. The experiment results demonstrate that BiGNN achieves outperforming performance with the average AUPROC of 0.947 and AUPR of 0.804. when evaluating on two benchmark datasets by 10-fold cross-validation.

Discovering new therapeutic indications for existing drugs is the essential part of drug repurposing. As a safe, low-cost and time-saving drug discovery technique, drug repurposing is attracting more and more research attention. However, there are still many potential drug-disease therapeutic effects uncovered in biological data sources. The advance of computation techniques offers great support for drug repurposing. In this study, we propose a computational approach, named BiGNN, to explore potential drug-disease associations. First, we collect multiple biological data sources to construct the heterogeneous information networks. Then BiGNN employs a bilevel graph-based neural network frameworks to aggregating features via node-level embedding and graph-level embedding respectively. Finally, the aggregated features are used to identify drug-disease associations. The experiment results demonstrate that BiGNN achieves outperforming performance with the average AUPROC of 0.95S and AUPR of 0.647 when evaluating on two benchmark datasets by 10-fold cross-validation.

Traditional drug development is a time-consuming, high-cost and high-risk process. Computational drug repositioning has become an important strategy to discover new indication for existing drugs due to its remarkable reduction in time, cost and risk. Related studies have indicated that data integration could be helpful to discover novel indications of drugs. However, how to adopt effective system framework to represent and integrate data from different data sources remains a challenging problem. In this study, we propose a computational approach named BioHNCFR, which employs a hybrid neural collaborative filtering framework combined with data integration to predict the potential drug indications. The experimental results on benchmark data set reveal that the performance of our model is improved in predict drug-diseases associations compared to the recent state-of-the-art approaches when evaluated by 10-fold cross validation. The AUC, AUPR on 10-fold cross validation, 93.24%, 39.89%. The HitRate@l reach to 45.54%. More comprehensive experiments show that our model can predict indications for new drugs.

Extracting adverse drug events (a.k.a. drug-induced diseases or drug side effects) from the raw text has been widely studied in the biomedical area. It is usually assumed that entities of drugs and diseases are given by a separated named entity recognition model. In this paper, we propose a joint model to extract drugs, diseases and adverse drug events simultaneously. In our model, the structured perceptron is leveraged for training and multiple-beam search algorithm is used for decoding. The search algorithm is transition-based, i.e., an input sentence is processed in left-to-right order and predefined actions transit the sentence from one state to another which corresponds to a predicted result about drugs, diseases and adverse drug events in that sentence. Experimental results show that our joint approach obtains comparable performance compared with the baseline or state-of-the-art approaches and achieves 55.20% precision, 47.97% recall and 51.14% F1-measure in extraction of adverse drug events. We demonstrate that the joint approach is effective and can be easily extended to other entity-relation extraction systems such as protein-protein interactions and gene-disease relations. To facilitate the related research, our code is available online at: https://github.com/foxlf823/ade.