As medical safety and drug regulation gain heightened attention, the detection of spurious drug-drug interactions (DDI) has become key in healthcare. Although current research using graph neural networks (GNNs) to predict DDI has shown impressive results, it often fails to account for false DDI in the constructed DDI networks. Such inaccuracies caused by data errors, false alarms, or incorrect drug details can skew the network's structure and hinder the accuracy of GNN-based predictions. To tackle this challenge, we propose ANSM, a network-enhancement method specifically designed to identify and attenuate spurious links between drugs for ensuring the accuracy of DDI networks. ANSM integrates three key components: the feature extractor, the network optimizer, and the discriminative classifier. The feature extractor captures local structural features from drug node pairs, while the network optimizer leverages network information to improve feature extraction and reduce the impact of spurious DDI links. The discriminative classifier then identifies potential spurious links. Experimental results demonstrate that ANSM outperforms state-of-the-art methods in identifying spurious DDI.

With the growing popularity of artificial intelligence in drug discovery, many deep-learning technologies have been used to automatically predict unknown drug-target interactions (DTIs). A unique challenge in using these technologies to predict DTI is fully exploiting the knowledge diversity across different interaction types, such as drug-drug, drug-target, drug-enzyme, drug-path, and drug-structure types. Unfortunately, existing methods tend to learn the specifical knowledge on each interaction type and they usually ignore the knowledge diversity across different interaction types. Therefore, we propose a multitype perception method (MPM) for DTI prediction by exploiting knowledge diversity across different link types. The method consists of two main components: a type perceptor and a multitype predictor. The type perceptor learns distinguished edge representations by retaining the specifical features across different interaction types; this maximizes the prediction performance for each interaction type. The multitype predictor calculates the type similarity between the type perceptor and predicted interactions, and the domain gate module is reconstructed to assign an adaptive weight to each type perceptor. Extensive experiments demonstrate that our proposed MPM outperforms the state-of-the-art methods in DTI prediction.

MOTIVATION: Bacteria can usually acquire certain advantageous genes that enable the bacteria to adapt to rapidly changing niches, thereby leading to a wide range of intraspecific genome content and genetic redundancy. The minimal genome of Escherichia coli, which is the most important bacterial species, and the association between E.coli and its human host are worthy of further exploration. RESULTS: We used gene prediction and phylogenetic analysis to reveal a rich phylogenetic diversity among 491 E.coli strains and to reveal substantial differences between these strains with respect to gene number and genome length. We used pan-genomic analysis to accurately identify 867 core genes, in which only 243 genes are shared by essential genes. This analysis revealed that core genes mainly provide essential functions to the basic lifestyle of E.coli, and accessory genes are likely to confer selective advantages such as niche adaptation or the ability to colonize specific hosts. By association analysis, we found that E.coli strains in non-human hosts may more easily utilize foreign genetic materials to adapt to their surroundings, but the population in human hosts has higher demands for the control of population density, indicating that highly accurate quorum-sensing behavior is very important for harmony between E.coli and its human host. By considering core genes and previous deletions together, we proposed a potential direction for further reduction of the E.coli genome. AVAILABILITY AND IMPLEMENTATION: The data, analysis process and detailed information on software tools used in this study are all available in the supplementary material. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Motivation: The budding yeast Saccharomyces cerevisiae is a model species powerful for studying the recombination of eukaryotes. Although many recombination studies have been performed for this species by experimental methods, the population genomic study based on bioinformatics analyses is urgently needed to greatly increase the range and accuracy of recombination detection. Here, we carry out the population genomic analysis of recombination in S.cerevisiae to reveal the potential rules between recombination and evolution in eukaryotes. Results: By population genomic analysis, we discover significantly more and longer recombination events in clinical strains, which indicates that adverse environmental conditions create an obviously wider range of genetic combination in response to the selective pressure. Based on the analysis of recombinational double strand breaks (DSBs)-intersected genes (RDIGs), we find that RDIGs significantly converge on specific disease- and adaptability-related pathways, indicating that recombination plays a biologically key role in the repair of DSBs related to diseases and environmental adaptability, especially the human neurological disorders. By evolutionary analysis of RDIGs, we find that the RDIGs highly prevailing in populations of yeast tend to be more evolutionarily conserved, indicating the accurate repair of DSBs in these RDIGs is critical to ensure the eukaryotic survival or fitness. Supplementary information: Supplementary data are available at Bioinformatics online.