Search

Deep learning has significantly advanced the field of pipeline circumferential guided wave testing. However, the scarcity of defective samples in non-destructive testing datasets poses a significant challenge due to data imbalance. Traditional computer vision augmentation techniques, such as rotation and flipping, are often not directly applicable to guided wave data due to its unique characteristics. To address this issue, we propose a novel approach that combines the strengths of convolutional neural networks (CNNs) and random forest (RF) classifiers. Raw detection data are preprocessed using continuous wavelet transform to generate informative time-frequency images. Random forests enhance predictive accuracy by combining multiple decision trees. The experiment results demonstrate the effectiveness of the proposed CNN-RF approach, particularly in handling imbalanced datasets. When the data imbalance ratio is more than 1/2, the shallow, lightweight CNN could effectively extract signal features and lead to approximately a 56.5% speedup. When the imbalance ratio is less than 1/2, the ResNet34-RF model significantly outperforms the standard CNN model. This approach holds potential for contributing to the miniaturization, lightweight design, and mobility of guided wave detection systems and devices.

Congenital heart disease (CHD) represents a spectrum of inborn heart defects influenced by genetic and environmental factors. This study advances the field by analyzing gene expression profiles in 21,034 cardiac fibroblasts, 73,296 cardiomyocytes, and 35,673 endothelial cells, utilizing single-cell level analysis and machine learning techniques. Six CHD conditions: dilated cardiomyopathy (DCM), donor hearts (used as healthy controls), hypertrophic cardiomyopathy (HCM), heart failure with hypoplastic left heart syndrome (HF_HLHS), Neonatal Hypoplastic Left Heart Syndrome (Neo_HLHS), and Tetralogy of Fallot (TOF), were investigated for each cardiac cell type. Each cell sample was represented by 29,266 gene features. These features were first analyzed by six feature-ranking algorithms, resulting in several feature lists. Then, these lists were fed into incremental feature selection, containing two classification algorithms, to extract essential gene features and classification rules and build efficient classifiers. The identified essential genes can be potential CHD markers in different cardiac cell types. For instance, the LASSO identified key genes specific to various heart cell types in CHD subtypes. FOXO3 was found to be up-regulated in cardiac fibroblasts for both Dilated and hypertrophic cardiomyopathy. In cardiomyocytes, distinct genes such as TMTC1, ART3, ARHGAP24, SHROOM3, and XIST were linked to dilated cardiomyopathy, Neo-Hypoplastic Left Heart Syndrome, hypertrophic cardiomyopathy, HF-Hypoplastic Left Heart Syndrome, and Tetralogy of Fallot, respectively. Endothelial cell analysis further revealed COL25A1, NFIB, and KLF7 as significant genes for dilated cardiomyopathy, hypertrophic cardiomyopathy, and Tetralogy of Fallot. LightGBM, Catboost, MCFS, RF, and XGBoost further delineated key genes for specific CHD subtypes, demonstrating the efficacy of machine learning in identifying CHD-specific genes. Additionally, this study developed quantitative rules for representing the gene expression patterns related to CHDs. This research underscores the potential of machine learning in unraveling the molecular complexities of CHD and establishes a foundation for future mechanism-based studies.

With the development of $3D$ visualization technology, the amount of geological data information is increasing, and the interactive display of big data faces severe challenges. Because traditional volume rendering methods cannot entirely load large-scale data into the memory owing to hardware limitations, a visualization method based on variational deep embedding clustering fusion Hilbert R-tree is proposed to solve slow display and stuttering issues when rendering massive geological data. By constructing an efficient data index structure, deep clustering algorithms and space-filling curves can be integrated into the data structure to improve the indexing efficiency. In addition, this method combines time forecasting, data scheduling, and loading modules to improve the accuracy and real-time data display rate, thereby improving the stability of $3D$ visualization of large-scale geological data. This method uses real geological data as the experimental dataset, comparing and analyzing the existing index structure and time-series prediction method. The experimental results indicate that when comparing the index of the variational deep embedded clustering-Hilbert R-tree ( $VDEC-HRT$ ) with that of the K-means Hilbert R-tree ( $KHRT$ ), the time required is reduced by 55.67%, the viewpoint prediction correctness of the proposed method is improved by 22.7% compared with Lagrange interpolation algorithm. And the overall rendering performance and quality of the system achieve the expected results. Ours experiments prove the feasibility and effectiveness of the proposed scheme in the visualization of large-scale geological data.

Innovation in facial recognition technology is urgent in the current epidemic situation. In order to extract more face features to perform facial recognition on images of faces with masks, we propose a multi-step iterative face-mask segmentation algorithm (MISA). The improved Mask R-CNN algorithm is used to detect the target in the face region to realize coarse segmentation, and the generalization ability and accuracy of the segmentation are improved. Then, the proposed approach uses an R-Pairwise Differential Siamese Network (R-PDSN) to train a mask occlusion classifier to subdivide the edge blocks of coarse segmentation results. The segmentation accuracy is further improved by optimizing the edge information of the masks step by step. The self-built dataset of faces with masks was used for training and testing. The experimental results showed that the mean pixel accuracy of the proposed method was improved by 2.69% compared with the original Mask R-CNN segmentation algorithm, and the target detection accuracy was more than 98%. These results indicate that the proposed method can achieve good segmentation performance on face images with complex backgrounds, self-occlusion and different types of masks. These results demonstrate that our method can improve the accuracy of segmentation methods for imaged of faces wearing masks.

Objectives. We aimed to establish an effective machine learning (ML) model for predicting the risk of distant metastasis (DM) in medullary thyroid carcinoma (MTC). Methods. Demographic data of MTC patients were extracted from the Surveillance, Epidemiology, and End Results (SEER) database of the National Institutes of Health between 2004 and 2015 to develop six ML algorithm models. Models were evaluated based on accuracy, precision, recall rate, F1-score, and area under the receiver operating characteristic curve (AUC). The association between clinicopathological characteristics and target variables was interpreted. Analyses were performed using traditional logistic regression (LR). Results. In total, 2049 patients were included and 138 developed DM. Multivariable LR showed that age, sex, tumor size, extrathyroidal extension, and lymph node metastasis were predictive features for DM in MTC. Among the six ML models, the random forest (RF) had the best predictability in assessing the risk of DM in MTC, with an accuracy, precision, recall rate, F1-score, and AUC higher than those of the traditional binary LR model. Conclusion. RF was superior to traditional LR in predicting the risk of DM in MTC and can provide a valuable reference for clinicians in decision-making.

Protein–protein interactions (PPIs) are extremely important for gaining mechanistic insights into the functional organization of the proteome. The resolution of PPI functions can help in the identification of novel diagnostic and therapeutic targets with medical utility, thus facilitating the development of new medications. However, the traditional methods for resolving PPI functions are mainly experimental methods, such as co-immunoprecipitation, pull-down assays, cross-linking, label transfer, and far-Western blot analysis, that are not only expensive but also time-consuming. In this study, we constructed an integrated feature selection scheme for the large-scale selection of the relevant functions of PPIs by using the Gene Ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotations of PPI participants. First, we encoded the proteins in each PPI with their gene ontologies and KEGG pathways. Then, the encoded protein features were refined as features of both positive and negative PPIs. Subsequently, Boruta was used for the initial filtering of features to obtain 5684 features. Three feature ranking algorithms, namely, least absolute shrinkage and selection operator, light gradient boosting machine, and max-relevance and min-redundancy, were applied to evaluate feature importance. Finally, the top-ranked features derived from multiple datasets were comprehensively evaluated, and the intersection of results mined by three feature ranking algorithms was taken to identify the features with high correlation with PPIs. Some functional terms were identified in our study, including cytokine–cytokine receptor interaction (hsa04060), intrinsic component of membrane (GO:0031224), and protein-binding biological process (GO:0005515). Our newly proposed integrated computational approach offers a novel perspective of the large-scale mining of biological functions linked to PPI.

Cervical and anal carcinoma are neoplastic diseases with various intraepithelial neoplasia stages. The underlying mechanisms for cancer initiation and progression have not been fully revealed. DNA methylation has been shown to be aberrantly regulated during tumorigenesis in anal and cervical carcinoma, revealing the important roles of DNA methylation signaling as a biomarker to distinguish cancer stages in clinics. In this research, several machine learning methods were used to analyze the methylation profiles on anal and cervical carcinoma samples, which were divided into three classes representing various stages of tumor progression. Advanced feature selection methods, including Boruta, LASSO, LightGBM, and MCFS, were used to select methylation features that are highly correlated with cancer progression. Some methylation probes including cg01550828 and its corresponding gene RNF168 have been reported to be associated with human papilloma virus-related anal cancer. As for biomarkers for cervical carcinoma, cg27012396 and its functional gene HDAC4 were confirmed to regulate the glycolysis and survival of hypoxic tumor cells in cervical carcinoma. Furthermore, we developed effective classifiers for identifying various tumor stages and derived classification rules that reflect the quantitative impact of methylation on tumorigenesis. The current study identified methylation signals associated with the development of cervical and anal carcinoma at qualitative and quantitative levels using advanced machine learning methods.

Mammalian cortical interneurons (CINs) could be classified into more than two dozen cell types that possess diverse electrophysiological and molecular characteristics, and participate in various essential biological processes in the human neural system. However, the mechanism to generate diversity in CINs remains controversial. This study aims to predict CIN diversity in mouse embryo by using single-cell transcriptomics and the machine learning methods. Data of 2,669 single-cell transcriptome sequencing results are employed. The 2,669 cells are classified into three categories, caudal ganglionic eminence (CGE) cells, dorsal medial ganglionic eminence (dMGE) cells, and ventral medial ganglionic eminence (vMGE) cells, corresponding to the three regions in the mouse subpallium where the cells are collected. Such transcriptomic profiles were first analyzed by the minimum redundancy and maximum relevance method. A feature list was obtained, which was further fed into the incremental feature selection, incorporating two classification algorithms (random forest and repeated incremental pruning to produce error reduction), to extract key genes and construct powerful classifiers and classification rules. The optimal classifier could achieve an MCC of 0.725, and category-specified prediction accuracies of 0.958, 0.760, and 0.737 for the CGE, dMGE, and vMGE cells, respectively. The related genes and rules may provide helpful information for deepening the understanding of CIN diversity.

Cancer driver gene is a type of gene with abnormal alterations that initiate or promote tumorigenesis. Driver genes can be used to reveal the fundamental pathological mechanisms of tumorigenesis. These genes may have pathological changes at different omics levels. Thus, identifying cancer driver genes involving two or more omics levels is essential. In this study, a computational investigation was conducted on lung cancer driver genes. Four omics levels, namely, epigenomics, genomics, transcriptomics, and post-transcriptomics, were involved. From the driver genes at each level, the Laplacian heat diffusion algorithm was executed on a protein–protein interaction network for discovering latent driver genes at this level. A following screen procedure was performed to extract essential driver genes, which contained three tests: permutation, association, and function tests, which can exclude false-positive genes and screen essential ones. Finally, the intersection operation was performed to obtain novel driver genes involving two omic levels. The analyses on obtained genes indicated that they were associated with fundamental pathological mechanisms of lung cancer at two corresponding omics levels.

Background: Intestinal Fusobacterium nucleatum (F. nucleatum) infection has been implicated into the progression of colorectal cancer (CRC). However, F. nucleatum as a biomarker in 5-fluorouracil (5-FU) resistance of CRC has not been fully analyzed by comparing with other types of gut microbiota. In order to further reduce the random error and bias of individual research, this meta‑analysis aimed to compare the diagnostic performance of intestinal Fusobacterium nucleatum, Bacteroides fragilis and Escherichia coli in 5-FU resistance to colorectal cancer and provide evidence‑based data to clinical practice. Methods: Comprehensive searches of PubMed, Embase, Cochrane Library and Web of Science databases were conducted. A total of 11 studies were selected according to the preestablished inclusion and exclusion criteria and analyzed by Review Manager 5.4 software, which included mapping of forest plots, heterogeneity tests, meta‑regression, sensitivity analysis and publication bias. Results: Overall sensitivity and specificity of F. nucleatum detection in 5-FU resistance of CRC were 0.65 (95% CI:0.60–0.69) and 0.70 (95% CI:0.59–0.87), respectively. Its positive likelihood ratio (PLR) and negative likelihood ratio (NLR) in detecting colorectal cancer were 2.57 (95% CI:1.47–3.21) and 0.52 (95% CI:0.43–0.63). The diagnostic odds ratio (DOR) was 4.92 (95% CI:2.23–7.33), which significantly exceeds the performance of B. fragilis (DOR: 0.53, 95% CI:0.31–0.82) and E. coli (DOR: 0.63, 95% CI: 0.57–0.76) for indicating 5-FU resistance of CRC. Conclusion: Compared with intestinal B. fragilis and intestinal E. coli, intestinal F. nucleatum is more reflective of 5-FU resistance to colorectal cancer.

Background: Thymosin-α1 has been implicated into the treatment of novel respiratory virus Coronavirus Disease 2019 (COVID-19), but the underlying mechanisms are still disputable. Aim: Herein we aimed to reveal a previously unrecognized mechanism that thymosin-α1 prevents COVID-19 by binding with angiotensin-converting enzyme (ACE), which was inspired from the tool of network pharmacology. Methods: KEGG pathway enrichment of thymosin-α1 treating COVID-19 was analyzed by Database of Functional Annotation Bioinformatics Microarray Analysis, then core targets were validated by ligand binding kinetics assay and fluorometric detection of ACE and ACE2 enzymatic activity. The production of angiotensin I, angiotensin II, angiotensin (1–7) and angiotensin (1–9) were detected by enzyme linked immunosorbent assay. Results: We found that thymosin-α1 impaired the expressions of angiotensin-converting enzyme 2 and angiotensin (1–7) of human lung epithelial cells in a dose-dependent way (p < 0.001). In contrast, thymosin-α1 had no impact on their ACE and angiotensin (1–9) expressions but significantly inhibited the enzymatic activity of ACE (p > 0.05). Conclusion: The bioinformatic findings of network pharmacology and the corresponding pharmacological validations have revealed that thymosin-α1 treatment could decrease ACE2 expression in human lung epithelial cells, which strengthens the potential clinical applications of thymosin-α1 to prevent severe acute respiratory syndrome coronavirus 2 infection.