In the process of pharmaceutical crystallization, the automatic detection of crystal shapes in images is important since controlling the morphology of the crystals improves the quality of pharmaceutical crystals. In this paper, a novel image detection method called RECDet is proposed. It leverages an automatically adapted binary image to bypass background regions, thereby reducing the detection field. In addition, the method greatly reduces the training time while improving the detection accuracy by using a specially designed detection box for the crystal shape. The performance of our model is evaluated through experimental analysis on a publicly available glutamate crystal dataset and a self-made entecavir pharmaceutical crystal dataset. Experimental results show that RECDet improves the accuracy of prediction bounding boxes by more than 2% compared to other popular models and achieves a classification accuracy of 98%. It can be used as a promising tool in the application of pharmaceutical crystallization control.
Introduction: Various drugs can markedly disrupt gut microbiota, resulting in a reduction of beneficial microbial populations and precipitating a range of negative clinical consequences. Traditional experimental methods have considerable limitations in clarifying the mechanisms of microbe-drug interactions, thereby necessitating the creation of innovative computational techniques to establish theoretical foundations for personalized and precision medicine. However, the majority of current computational methods rely on graph structures, which inadequately represent the intricate, varied, and heterogeneous interactions among multiple drugs and microbial communities.Methods: We introduce a hierarchical attention-driven dual-hypergraph contrastive learning framework for predicting microbe-drug interactions. Initially, the original bipartite graph and various similarity data are integrated using nonlinear features by incorporating the functional similarity of medicinal chemical attributes and microbial genomes, alongside computing the Gaussian kernel similarity. Subsequently, a dual network structure comprising K-Nearest Neighbors (KNN) hypergraph and K-means Optimizer (KO) hypergraph is established, employing a hierarchical attention mechanism to facilitate collaborative information aggregation between hyperedges and hypernodes. A contrastive learning approach is implemented to enhance the representation of the heterogeneous hypergraph space, and the prediction scores for microbe-drug interactions are derived by dynamically integrating two-channel embedded features via multi-head attention.Results: Experiments conducted on various publicly accessible benchmark datasets demonstrate that the DHCLHAM model markedly surpasses the current optimal model in critical metrics, including AUC and AUPR. Particularly on the aBiofilm dataset, the AUC and AUPR attained 98.61% and 98.33%, respectively.Discussion: A computational framework was developed through multi-dimensional case validation, integrating artificial intelligence and network pharmacology principles, offering a novel paradigm for analyzing microbe-drug interaction mechanisms. The research findings hold significant reference value for optimizing clinical treatment protocols and establish a theoretical foundation to develop precise medication strategies aimed at intestinal flora.
Virus infestation can seriously harm the host plant’s growth and development. Turnip yellows virus (TuYV) infestation of host plants can cause symptoms, such as yellowing and curling of leaves and root chlorosis. However, the regulatory mechanisms by which TuYV affects host growth and development are unclear. Hence, it is essential to mine small RNA (sRNA) and explore the regulation of sRNAs on plant hosts for disease control. In this study, we analyzed high-throughput data before and after TuYV infestation in Arabidopsis using combined genetics, statistics, and machine learning to identify 108 specifically expressed and critical functional sRNAs after TuYV infection. First, comparing the expression levels of sRNAs before and after infestation, 508 specific sRNAs were significantly up-regulated in Arabidopsis after infestation. In addition, the results show that AI models, including SVM, RF, XGBoost, and CNN using two-dimensional convolution, have robust classification features at the sequence level, with a prediction accuracy of about 96.8%. A comparison of specific sRNAs with genome sequences revealed that 247 matched precisely with the TuYV genome sequence but not with the Arabidopsis genome, suggesting that TuYV viruses may be their source. The 247 sRNAs predicted target genes and enrichment analysis, which identified 206 Arabidopsis genes involved in nine biological processes and three KEGG pathways associated with plant growth and viral stress tolerance, corresponding to 108 sRNAs. These findings provide a reference for studying sRNA-mediated interactions in pathogen infection and are essential for establishing a vital resource of regulation network for the virus infecting plants and deepening the understanding of TuYV virus infection patterns. However, further validation of these sRNAs is needed to gain a new understanding.
The upsampling of point clouds is a common task to increase the expressiveness and richness of the details. The quality of upsampled point clouds is crucial for downstream tasks, such as mesh reconstruction. With the rapid development of deep learning technology, many neural network-based methods have been proposed for point cloud upsampling. However, there are common challenges among these methods such as blurring sharper points (e.g., corner or edge points) and producing points gathered together. These problems are caused by similar feature replication or insufficient supervised information. To address these concerns, we present SSPU-FENet, a self-supervised network consisting of two modules specifically designed for geometric detail-preserved point cloud upsampling. The first module, called the feature enhancement module (FEM), aims to prevent feature blurring. This module retains important features such as edges and corners by using non-artificial encoding methods and learning mechanisms to avoid the creation of blurred points. The second module, called the 3D noise perturbation module (NPM), focuses on high-dimensional feature processing and addresses the challenges of feature similarity. This module adjusts the spacing of reconstructed points, ensuring that they are neither too close nor too far apart, thus maintaining point uniformity. In addition, SSPU-FENet proposes self-supervised loss functions that emphasize global shape consistency and local geometric structure consistency. These loss functions enable efficient network training, leading to superior upsampling results. Experimental results on various datasets show that the upsampling results of the SSPU-FENet are comparable to those of supervised learning methods and close to the ground truth (GT) point clouds. Furthermore, our evaluation metrics, such as the chamfer distance (CD, 0.0991), outperform the best methods (CD, 0.0998) in the case of upsampling with 2048-point input.