KGE-UNIT: toward the unification of molecular interactions prediction based on knowledge graph and multi-task learning on drug discovery

Artificial intelligence to predict inhibitors of drug-metabolizing enzymes and transporters for safer drug design

INTRODUCTION

Drug-metabolizing enzymes (DMEs) and transporters (DTs) play integral roles in drug metabolism and drug-drug interactions (DDIs) which directly impact drug efficacy and safety. It is well-established that inhibition of DMEs and DTs often leads to adverse drug reactions (ADRs) and therapeutic failure. As such, early prediction of such inhibitors is vital in drug development. In this context, the limitations of the traditional in vitro assays and QSAR models methods have been addressed by harnessing artificial intelligence (AI) techniques.

AREAS COVERED

This narrative review presents the insights gained from the application of AI for predicting DME and DT inhibitors over the past decade. Several case studies demonstrate successful AI applications in enzyme-transporter interaction prediction, and the authors discuss workflows for integrating these predictions into drug design and regulatory frameworks.

EXPERT OPINION

The application of AI in predicting DME and DT inhibitors has demonstrated significant potential toward enhancing drug safety and effectiveness. However, critical challenges involve the data quality, biases, and model transparency. The availability of diverse, high-quality datasets alongside the integration of pharmacokinetic and genomic data are essential. Lastly, the collaboration among computational scientists, pharmacologists, and regulatory bodies is pyramidal in tailoring AI tools for personalized medicine and safer drug development.

Predicting genes associated with ossification of the posterior longitudinal ligament using graph attention network

Ossification of the posterior longitudinal ligament is a degenerative disease that severely impacts the spine, with a complex pathogenesis involving the interplay of multiple genes. This study utilizes a combination of graph neural networks and deep neural networks to systematically analyze genes associated with OPLL, leveraging genomics and bioinformatics techniques. By integrating gene data from the DisGeNET and HumanNetV2 databases, we constructed a GNN model to identify potential pathogenic genes for OPLL and validated the expression characteristics and mechanisms of these genes in different cell types. The findings indicate that the GNN model achieves remarkable accuracy and reliability in predicting genes associated with OPLL. Additionally, cellular trajectory analysis and immune cell infiltration studies uncovered distinct cellular environments and immune features in OPLL patients, emphasizing the significant roles of fibroblasts and mesenchymal stem cells in the disease's progression. Drug sensitivity analysis also sheds light on future personalized treatment options. This study not only enhances the understanding of OPLL's molecular mechanisms but also suggests new avenues for diagnostic and targeted therapy development.

KGRDR: a deep learning model based on knowledge graph and graph regularized integration for drug repositioning

Computational drug repositioning, serving as an effective alternative to traditional drug discovery plays a key role in optimizing drug development. This approach can accelerate the development of new therapeutic options while reducing costs and mitigating risks. In this study, we propose a novel deep learning-based framework KGRDR containing multi-similarity integration and knowledge graph learning to predict potential drug-disease interactions. Specifically, a graph regularized approach is applied to integrate multiple drug and disease similarity information, which can effectively eliminate noise data and obtain integrated similarity features of drugs and diseases. Then, topological feature representations of drugs and diseases are learned from constructed biomedical knowledge graphs (KGs) which encompasses known drug-related and disease-related interactions. Next, the similarity features and topological features are fused by utilizing an attention-based feature fusion method. Finally, drug-disease associations are predicted using the graph convolutional network. Experimental results demonstrate that KGRDR achieves better performance when compared with the state-of-the-art drug-disease prediction methods. Moreover, case study results further validate the effectiveness of KGRDR in predicting novel drug-disease interactions.

Drug-target prediction through self supervised learning with dual task ensemble approach

Drug-Target interaction (DTI) prediction, a transformative approach in pharmaceutical research, seeks novel therapeutic applications for computational method based virtual screening, existing drugs to address untreated diseases and discovery of existing drugs side effects. The proposed model predict DTI through Heterogeneous biological network by combining drug, genes and disease related knowledge. For the purpose of embedding extraction Self-supervised learning (SSL) has been used which, trains models through pretext tasks, eliminating the need for manual annotations. The pretext tasks are related to either structural based information or similarity based information. To mitigate GNN vulnerability to non-robustness, ensemble learning can be incorporated into GNNs, harnessing multiple models to enhance robustness. This paper introduces a Graph neural network based architecture consisting of task based module and ensemble module for link prediction of DTI. The ensemble module of dual task combinations, both in cold start and warm start scenarios achieve very good performance as it provide 0.960 in cold start and 0.970 in warm start mean AUCROC score with less deviation.