Multiple Machines Research Articles

Forward-model-based grain reconstruction to improve the tolerance of diffraction contrast tomography for increased sample deformation

To extend the applicability of synchrotron diffraction contrast tomography (DCT) towards more plastically deformed materials, we have developed a forward-model-based reconstruction method. This allows us to reconstruct grain shapes and local orientations in materials exhibiting levels of intragranular orientation spread that can no longer be handled with the conventional Friedel-pair-based indexing and tomographic reconstruction approach. This method consists of seed and regional indexing, in which an exhaustive searching and fitting of orientations is first performed to index the seed orientation, and then a regional indexing by testing a list of local orientations around the seed orientation is carried out to maximize the completeness. The capability of this novel method was benchmarked and compared with reconstructions based on the conventional Friedel-pair-matching and tomographic reconstruction method using samples made from fully recrystallized Al–Cu alloy, moderately deformed α-Ti alloy and 10% creep-ruptured Fe–Au alloy. The results show that this method has the potential to overcome the deformation constraint and can reconstruct reasonably well the intragranular orientations. It is also suitable for multi-phase reconstruction and both box-beam and line-beam acquisition geometries. The implementation has been made flexible to support the use of single or multiple GPU machines. The strengths and weaknesses of the current forward-model-based reconstruction are discussed in detail with respect to the conventional Friedel-pair-matching method. To fully exploit and complement the strengths of the two methods, the code to implement the current forward-model-based reconstruction has been fully integrated with the existing DCT code and is open source for beamline data processing.

ABIDS-VEM: leveraging an equilibrium optimizer and data ramification in association with ensemble learning for anomaly-based intrusion detection system

The convergence of the Internet of Things (IoT) and Industrial Internet of Things (IIoT) within the Industry 4.0 paradigm leverages software-defined networking, multi-cloud architectures, and edge/fog computing to enhance industrial processes. However, this digital transformation introduces significant cybersecurity and privacy vulnerabilities within the complex, data-intensive IoT/IIoT ecosystems. To mitigate these risks, this research proposes a novel Anomaly-based Intrusion Detection System using Voting-based Ensemble Model (ABIDS-VEM) in Industry 4.0 environments. The VEM architecture synergistically combines multiple machine learning algorithms and gradient boosting frameworks, including CatBoost (CB), XGBoost (XGB), LightGBM (LGBM), Logistic Regression (LR), and Random Forest (RF), to enhance the precision and computational efficiency of intrusion detection systems (IDS) in IoT/IIoT contexts. The proposed framework incorporates a data ramification process, in which the data is divided into multiple parts, feature selection process which is optimized through the Equilibrium Optimizer (EO) algorithm, and outlier detection utilizing the Isolation Forest (IF) method. Comprehensive empirical evaluations were conducted using three benchmark datasets: XIIoTID, NSL-KDD, and UNSW-NB15, to validate the efficacy of the proposed system. The model achieves high accuracy across datasets: 98.1476% for XIIoT-ID, an impressive accuracy of 98.9671% for NSL-KDD, and 94.1327% for UNSW-NB15 dataset. These experimental results demonstrate the potential of this approach to significantly enhance the resilience of critical industrial systems and data against evolving cyber threats, thereby supporting the continued evolution of Industry 4.0 technologies and bolstering the security posture of IoT/IIoT ecosystems. This research contributes to the ongoing efforts to secure the rapidly expanding digital industrial landscape, offering a robust solution for detecting and mitigating sophisticated cyberattacks in the increasingly interconnected and data-driven industrial environments of the future.

Smart quality control: integrating six sigma, machine learning and real-time defect prediction in manufacturing

Purpose This study aims to develop a smart quality control framework integrating Six Sigma methodology, real-time sensing technologies and machine learning algorithms to enhance manufacturing defect prediction and process optimization. By leveraging predictive capabilities and real-time data analysis, the framework seeks to reduce costs associated with poor quality and improve overall process capability. Design/methodology/approach The proposed framework uses the Define-Measure-Analyze-Improve-Control methodology to identify and address critical process parameters. A robust Internet of Things (IoT) sensing network is incorporated for continuous process monitoring. At the same time, multiple machines learning models, including decision trees, random forests, boosted decision trees, linear regression and k-star algorithms, are evaluated for predictive defect detection. Implementation was conducted at an electrical conductor manufacturing facility, enabling real-time analysis and intervention to prevent defects. Findings The implementation of the framework demonstrated significant improvements in quality and efficiency. The cost of poor quality was reduced from 5% to 1.7%, a 66% improvement. Process capability was enhanced, with sigma levels increasing from 3.14 to 4.3. These results validate the effectiveness of combining traditional quality control techniques with advanced Artificial Intelligence and IoT technologies, delivering predictive capabilities and enabling real-time process optimization. Originality/value This study highlights the innovative integration of Six Sigma, machine learning and IoT sensing technologies to transform manufacturing quality control. The smart quality control framework represents a significant advancement in manufacturing intelligence, offering a scalable, data-driven solution that improves efficiency, competitiveness and sustainability across diverse industrial applications.

Identification of novel potential biomarkers using bulk RNA and single cells to build a neural network model for diagnosis of liver cancer

BackgroundAs a common cancer, liver cancer imposes an unacceptable burden on patients, but its underlying molecular mechanisms are still not fully understood. Therefore, there is an urgent need to potential biomarkers and diagnostic models for liver cancer.MethodsIn this study, transcriptome and single-cell datasets related to liver cancer were downloaded from the UCSC Xena database and the Mendeley database, and differential analysis and weighted gene co-expression network analysis were used to find differentially expressed genes related to liver cancer. We used multiple machine algorithms to find hub genes related to liver cancer, and constructed new artificial neural network models based on their transcriptome expression patterns to assist in the diagnosis of liver cancer. Subsequently, we conducted survival analysis and immune infiltration analysis to explore the correlation between hub genes and immune cells, and used single-cell data to verify hub genes related to liver cancer.ResultsThis study identified MARCO, KCNN2, NTS, TERT and SFRP4 as central genes associated with liver cancer, and constructed a new artificial neural network model for molecular diagnosis of liver cancer. The diagnostic performance of the training cohort and the validation cohort was good, with the areas under the ROC curves of 1.000 and 0.986, respectively. Immune infiltration analysis determined that these central genes were closely associated with different types of immune cells. The results of immunohistochemistry and the results at the single cell level were consistent with those at the transcriptome level, and also showed obvious differences between different cell types in liver cancer and healthy states.ConclusionThis study identified MARCO, KCNN2, NTS, TERT, and SFRP4 from multiple dimensions and highlighted their key roles in the diagnosis and treatment of liver cancer from multiple dimensions, providing promising biomarkers for the diagnosis of liver cancer.

Development of a prognostic model for osteosarcoma based on macrophage polarization-related genes using machine learning: implications for personalized therapy

While neoadjuvant chemotherapy combined with surgical resection has improved the prognosis for patients with osteosarcoma, its impact on metastatic and recurrent cases remains limited. Immunotherapy is emerging as a promising alternative. However, the relationship between the phenotype of tumor-associated macrophages and the prognosis of osteosarcoma remains unclear. Differentially expressed gene during macrophage polarization were identified using the Monocle package. Weighted gene co-expression network analysis was conducted to select genes regulating macrophage polarization. The least absolute shrinkage and selection operator algorithm and multivariate Cox regression were used to construct long-term survival predictive strategies. Multiple machine learning algorithms identified target genes for pan-cancer analysis. Lentiviral transfection created stable strains with target gene knockdown, and CCK-8 and transwell migration assays verified the target gene's effects. Western blot and flow cytometry assessed the impact of target genes on macrophage polarization. A total of 141 genes regulating macrophage polarization were identified, from which eight genes were selected to construct prognostic models. Significant differences between high-risk and low-risk groups were observed in immune cell activation, immune-related signaling pathways, and immune function. The prognostic model and target gene were validated to provide more precise immunotherapy options for osteosarcoma and other tumors. BNIP3 knockdown decreased osteosarcoma cell proliferation and migration and promoted macrophage polarization to the M2 phenotype. The constructed prognostic model offers precise immunotherapy regimens and valuable insights into mechanisms underlying current studies. Furthermore, BNIP3 may serve as a potential immunotherapeutic target for osteosarcoma and other tumors.

Multiple machine learning algorithms identify 13 types of cell death-critical genes in large and multiple non-alcoholic steatohepatitis cohorts

BackgroundDysregulated programmed cell death pathways mechanistically contribute to hepatic inflammation and fibrogenesis in non-alcoholic steatohepatitis (NASH). Identification of cell death genes may offer insights into diagnostic and therapeutic strategies for NASH.MethodsData from multiple NASH cohorts were integrated, and 12 machine learning algorithms were applied to identify key dysregulated cell death-related genes and develop a binary classification model for NASH. Spearman's rank correlation coefficients quantified associations between these genes and clinical markers, immune infiltration profiles, and signature genes encoding pro-inflammatory mediators, metabolic regulators, and fibrotic drivers. Gene set enrichment analysis (GSEA) was performed to delineate the mechanistic underpinnings of these key genes. Consensus clustering analysis was then used to stratify patients with NASH into distinct phenotypic subgroups based on expression levels of these genes.ResultsA NASH prediction model, developed using the random forest (RF) algorithm, demonstrated high diagnostic accuracy across multiple cohorts. Four key genes, enriched in lipid metabolism and inflammation pathways, were identified. Their transcriptional levels were significantly correlated with the non-alcoholic fatty liver disease activity score (NAS), hepatic inflammatory infiltration, molecular signatures of metabolic dysregulation (lipid homeostasis regulators), and fibrosis progression. These genes also enabled accurate classification of patients with NASH into clusters reflecting varying disease severity.ConclusionsA binary classification model, developed using the RF algorithm, accurately identified patients with NASH. The four cell death genes, identified through 12 machine learning algorithms, represent potential biomarkers and therapeutic targets for NASH. These genes contribute to inflammation-related immune cell activation, lipid metabolism dysregulation, and liver fibrosis, highlighting the complex interplay between cell death and NASH progression.

Development and Validation of Interpretable Machine Learning Models Incorporating Paraspinal Muscle Quality to Predict Cage Subsidence Risk Following Posterior Lumbar Interbody Fusion.

A real-world, multicenter retrospective study. To identify independent risk factors for cage subsidence following Posterior Lumbar Interbody Fusion (PLIF) and develop an interpretable machine learning model for risk prediction. Patients with degenerative lumbar disease who underwent single-level PLIF (January 2018-October 2023) were retrospectively included. A training set (n=620) came from the First Affiliated Hospital of Soochow University, and a validation set (n=100) from the Second Affiliated Hospital. Cage subsidence (≥2mm intervertebral height loss) was assessed radiographically. Parameters included paraspinal muscle indices (Psoas Muscle Index [PMI], Multifidus Muscle Index [MM]), Fat Infiltration [FI]), bone density markers (Hounsfield Unit [HU] value, Vertebral Bone Quality [VBQ], Endplate Bone Quality [EBQ]), cage position, and postoperative alignment. Multivariate logistic regression identified risk factors; multiple machine learning models were developed and evaluated. A web-based tool was created for clinical deployment. Multivariate analysis identified PMI, FI, HU value, VBQ, cage position, cage height, postoperative Intervertebral Height (IH), corrected IH, and corrected SA as independent risk factors for cage subsidence. Light Gradient Boosting Machine (LightGBM) outperformed other models, achieving the highest AUC (0.9752), accuracy (0.92), and F1-score (0.9216), with the lowest Brier score (0.0660). After excluding indicators related to paravertebral muscle function from the prediction model, the predictive accuracy of the model decreased substantially. (SHapley Additive exPlanations) SHAP analysis confirmed VBQ, PMI, BMI, and EBQ as the most influential predictors. The final model was deployed as a web-based tool for real-time clinical risk assessment. Key risk factors for PLIF cage subsidence were identified, and a validated machine learning model was developed. The high-performance LightGBM model, deployed in a user-friendly web application, enables spine surgeons to optimize surgical planning and reduce subsidence risk.

An ensemble model with convolutional neural network by DS evidence fusion for bearing fault diagnosis

Bearing fault diagnosis is crucial for ensuring the safety and reliability of rotating machinery. In recent years, artificial intelligence technology based on machine learning has made substantial progress in the field of bearing fault diagnosis. Most existing models for bearing fault diagnosis are built using big data and deep learning algorithms and can achieve high diagnostic accuracy with sufficient fault data. However, there still exist two open issues, 1) in practical engineering, acquiring fault sample data is challenging, and it is difficult to obtain a sufficient number of samples to train the hyperparameters of deep learning models. 2) Fault diagnosis models based on individual classifiers rely heavily on prior knowledge for signal feature extraction and the selection of network structures and parameters, making it difficult to guarantee the model’s effectiveness. This paper proposes an integrated diagnostic model called DS-ELM that employs multiple extreme learning machine modules with different parameters as subclassifiers. The outputs of these modules are then fused via DS evidence fusion theory to obtain the final diagnostic result. This ensemble model has better flexibility and robustness which significantly improves the accuracy and stability of the diagnostic model. Overall, the proposed DS-ELM provides a new solution for bearing fault diagnosis. In addition, the superiority of the reported technique is confirmed via experimental bearing fault data from Case Western Reserve University.

Influences of Sampling Design and Model Selection on Predictions of Chemical Compounds in Petroferric Formations in the Brazilian Amazon

Morro de Seis Lagos, a region in the Brazilian Amazon, contains a small (less than 1%) formation of siderite carbonatites which is considered to be one of the world’s largest niobium reserves. This highly weathered geological and pedological occurrence makes the site ideal for studying the pedogenetic process of lateralization and the spatial variability of chemical elements. The aim of this study was to investigate the influences of various sampling combinations (scenarios) derived from three sampling designs on the spatial predictions associated with chemical compounds (Al2O3, Fe2O3, MnO, Nb2O5, TiO2, and SiO2), using multiple machine learning algorithms combined with remotely sensed imagery. The dataset comprised 341 samples from the Geological Survey of Brazil (CPRM). Covariates included remotely sensed data collected from Sentinel-2 MSI, Sentinel-1A, and the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), and topographic attributes were calculated from a 20 m digital elevation model derived from hydrologic data (HC-DEM). The machine learning algorithms (Generalized Linear Models with Elastic Net Regularization (GLMNET), Nearest Neighbors (KNN), Neural Network (NNET), Random Forest (RF) and Support Vector Machine (SVMRadial) were used in combination with covariates and measured elements at point locations to spatially map the concentrations of these chemical elements. The optimal covariates for modeling were selected using Recursive Feature Elimination (RFE), processing 10 runs for each chemical element. The RF, SVMRadial, and KNN models performed best, followed by the models from the Neural Network group (NNET). The sampling scenarios were not significantly different, based on root mean square error (F = 1.7; p-value = 0.15) and mean absolute error (F = 0.4; p-value = 0.79); however, significant differences were observed in the coefficient of determination (F = 41.2; p-value < 0.00) across all models. Overall, the models performed poorly for all elements, with R2 ranging from 0.07 to 0.27, regardless of sampling scenario (F = 1.6; p-value = 0.08). Relatively, RF, GLMET, and KNN performed better, compared to other models. The terrain attributes were significantly more successful as to the spatial predictions of the elements contained in laterites than were the remote sensing spectral indices, likely due to the fact that the underlying spatial structures of the two formations (laterite and talus) occur at different elevations.

Multi-Domain Controversial Text Detection Based on a Machine Learning and Deep Learning Stacked Ensemble

Due to the rapid proliferation of social media and online reviews, the accurate identification and classification of controversial texts has emerged as a significant challenge in the field of natural language processing. However, traditional text-classification methodologies frequently encounter critical limitations, such as feature sensitivity and inadequate generalization capabilities. This results in a notably suboptimal performance when confronted with diverse controversial content. To address these substantial limitations, this paper proposes a novel controversial text-detection framework based on stacked ensemble learning to enhance the accuracy and robustness of text classification. Firstly, considering the multidimensional complexity of textual features, we integrate comprehensive feature engineering, i.e., encompassing word frequency, statistical metrics, sentiment analysis, and comment tree structure features, as well as advanced feature selection methodologies, particularly lassonet, i.e., a neural network with feature sparsity, to effectively address dimensionality challenges while enhancing model interpretability and computational efficiency. Secondly, we design a two-tier stacked ensemble architecture, which not only combines the strengths of multiple machine learning algorithms, e.g., gradient-boosted decision tree (GBDT), random forest (RF), and extreme gradient boosting (XGBoost), with deep learning models, e.g., gated recurrent unit (GRU) and long short-term memory (LSTM), but also implements the support vector machine (SVM) for efficient meta-learning. Furthermore, we systematically compare three hyperparameter optimization algorithms, including the sparrow search algorithm (SSA), particle swarm optimization (PSO), and Bayesian optimization (BO). The experimental results demonstrate that the SSA exhibits a superior performance in exploring high-dimensional parameter spaces. Extensive experimentation across diverse topics and domains also confirms that our proposed methodology significantly outperforms the state-of-the-art approaches.

Disease Prediction and Medication Recommendation Using Machine Learning

Abstract—Advancements in healthcare have greatly benefited from machine learning, especially in disease prediction and medication recommendations. This project introduces a Python-based machine learning model designed to predict potential diseases based on user-reported symptoms and suggest suitable medications. We implemented and evaluated four machine learning algorithms—Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and Naïve Bayes—to determine the most accurate model for deployment, ensuring reliable predictions. To make the system user-friendly, we developed an interactive graphical interface using Tkinter. This allows users to easily input their symptoms and receive potential diagnoses, along with relevant precautions and medication suggestions. The recommendation system identifies appropriate pharmaceutical salts for the diagnosed condition, improving medication accuracy and guidance. By comparing multiple machine learning models, our approach ensures that the most accurate algorithm is selected, enhancing the reliability of disease predictions. The intuitive GUI makes healthcare support more accessible, even for individuals without medical expertise, bridging the gap between users and early healthcare assessments. Future improvements will focus on expanding the dataset, integrating deep learning techniques for better accuracy, and connecting with real-time medical databases to provide up-to-date medication recommendations. This project highlights the potential of artificial intelligence in transforming early disease detection and medication guidance, ultimately improving healthcare accessibility and decision-making. Keywords—Machine Learning, Disease Prediction, Medication Recommendation, Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Naïve Bayes, Graphical User Interface (GUI), Healthcare Accessibility, Artificial Intelligence.

Correlation analysis between warpage deformation and optical sensing signals during LPBF process

Abstract Laser Powder Bed Fusion (LPBF) technology exhibits the potential to revolutionize traditional manufacturing processes. However, manufacturing repeatability and quality assurance have yet to be optimized. A pivotal obstacle is the lack of effective real-time monitoring, which is crucial for reducing defect incidence and maintaining melt pool stability. Furthermore, a comprehensive analysis that combines real-time monitoring signals, process parameters, and part quality is essential for effective monitoring. Therefore, a suite of test experiments was conducted using standard overhanging specimens, varying support column dimensions and scanning strategies. In-situ monitoring techniques were employed to provide an in-depth investigation of the interactions between optical sensing signals, specimen warpage deformation and process parameters. And based on the correlation analysis among the three, this paper developed a method to qualitatively predict specimen warpage. Finally, this study validated and evaluated the proposed warpage deformation classification method using multiple machine learning models. The primary findings of the analysis indicate that the light intensity signal from the initial overhanging layer captures critical process information that is essential for specimen fabrication and significantly affects the light intensity signal of subsequent layers. The overall light intensity signals of the overhanging layers exhibit a trend of increasing initially and then stabilizing as the number of sintered layers grows. From the smart island scanning group to the normal island scanning group, and subsequently to the zigzag scanning group, a consistent increase in warpage deformation is observed. An inverse correlation exists between the dimensions of support columns and the warpage deformation of overhanging specimens in zigzag scanning.

Abnormal Sound Source Detection and Localization by Spatial Mapping of Normal Sounds for Robotic Inspection in Oil Refineries

In this paper, we propose a novel approach to detection and localization of abnormal sound sources for robotic inspection in oil refineries. Such environments are difficult environments with high noise from multiple machines and where swift detection of anomalies is critical. The rarity of anomalies hinders the gathering of a balanced training dataset for the common supervised learning approach. Our previous work, based on autoencoders, bypassed this issue but lacked the ability to locate the abnormal sound source. Our proposed method first learns a spatial map of the normal sounds, allowing to predict what sound should be present at each robot position. This enables a detection based on a comparison between the predicted and observed sound. Localization can then be conducted based on this comparison using optimization. Experiments conducted in laboratory conditions showed the effectiveness of the proposed method. Additionally, experiments in field conditions in an actual oil refinery further showed the potential of the proposed method.

An Efficient Multiple Empirical Kernel Learning Algorithm with Data Distribution Estimation

The Multiple Random Empirical Kernel Learning Machine (MREKLM) typically generates multiple empirical feature spaces by selecting a limited group of samples, which helps reduce training duration. However, MREKLM does not incorporate data distribution information during the projection process, leading to inconsistent performance and issues with reproducibility. To address this limitation, we introduce a within-class scatter matrix that leverages the distribution of samples, resulting in the development of the Fast Multiple Empirical Kernel Learning Incorporating Data Distribution Information (FMEKL-DDI). This approach enables the algorithm to incorporate sample distribution data during projection, improving the decision boundary and enhancing classification accuracy. To further minimize sample selection time, we employ a border point selection technique utilizing locality-sensitive hashing (BPLSH), which helps in efficiently picking samples for feature space development. The experimental results from various datasets demonstrate that FMEKL-DDI significantly improves classification accuracy while reducing training duration, thereby providing a more efficient approach with strong generalization performance.

Screening of pivotal oncogenes modulated by DNA methylation in hepatocellular carcinoma and identification of atractylenolide I as an anti-cancer drug.

This study was performed to identify crucial oncogenes modulated by DNA methylation in hepatocellular carcinoma (HCC) and look for new drugs for HCC treatment. The data of TCGA-LIHC cohort were obtained from UCSC database. Weighted gene co-expression network analysis and multiple machine learning algorithms were applied to screen the crucial prognosis-related genes in HCC. Then these genes were further screened by DNA methylation status. Ten-eleven translocation 1 (TET1) was overexpressed in HCC cell lines, and its biological functions and regulatory effects on the oncogenes were explored by qPCR, methylation-specific polymerase chain reaction, cell viability assay, Western blot, etc. Molecular docking was applied to evaluate the binding affinity between atractylenolide I (AT-I) and TET1, and the tumor-suppressive functions of AT-I were examined with both in vitro and in vivo models. In this work, 12 crucial genes related to HCC prognosis were obtained, among which six genes were with differential methylation status in HCC tissues, including AKR1B10, ALPK3, NQO1, NT5DC2, SFN, and SPP1. The expression levels of ALPK3 and NT5DC2 were positively regulated by TET1, the crucial mediator of demethylation. TET1 overexpression increased the viability and stemness of HCC cells. AT-I had good binding affinity with TET1, and repressed its activity. AT-I promoted the methylation of ALPK3 and NT5DC2 promoter regions, and reduced their expression, and repressed the growth of HCC cells. In summary, DNA methylation contributes to HCC progression, and AT-I represses the malignancy of HCC cells by inhibiting TET1-mediated abnormal DNA demethylation.

The immunosuppressive role of VSIG4 in colorectal cancer and its interaction with the tumor microenvironment

BackgroundThe tumor microenvironment in colorectal cancer (CRC) significantly influences disease progression and immune responses, particularly the role of macrophages in regulating immune evasion requires further investigation.MethodsThis study integrated data from the TCGA-COAD dataset with the GEO database, along with single-cell RNA sequencing data, to systematically analyze key genes in colorectal cancer. R software was utilized for data normalization and differential analysis, with criteria set at ∣log2FoldChange ∣ > 1 and adjusted p-value < 0.05 for gene selection. The Seurat package was employed for clustering single-cell data, while the “Monocle2” algorithm was used to perform pseudo-time analysis on the differentiation trajectory of macrophages. Additionally, non-negative matrix factorization (NMF) was applied for subtype classification of CRC patients, and various machine learning algorithms (such as LASSO and random forest models) were utilized to identify key pathogenic genes. Finally, PCR was employed to validate the expression of these key genes, and immune analysis software was used to assess their impact on immune cells, alongside pathway enrichment analysis.ResultsThrough the integration of multi-omics data, we identified significant differential expression of VSIG4, CYBBC3AR1, and FCGR1A in CRC patients. LASSO and random forest models selected these three genes as critical pathogenic factors for CRC, with AUC values exceeding 0.8 across multiple machine learning models, demonstrating their high diagnostic efficacy. PCR validation further supported the differential expression of VSIG4 and other genes in CRC. Single-cell transcriptomic analysis revealed that VSIG4 was highly enriched in specific macrophage subpopulations and significantly influenced the tumor microenvironment by regulating CD8 + T cell immune exhaustion. Pseudo-time analysis indicated that the differentiation trajectory of macrophages during tumor progression was closely associated with VSIG4 expression. Additionally, cell communication analysis.highlighted the important role of VSIG4 in the interactions between macrophages and endothelial cells. Pathway enrichment analysis demonstrated that VSIG4 expression was closely linked to the regulation of the JAK-STAT pathway and metabolic pathways such as the TCA cycle.ConclusionThis study provides the first evidence that VSIG4, CYBBC3AR1, and FCGR1A play critical roles in the immune microenvironment of colorectal cancer, particularly emphasizing the immunoregulatory function of VSIG4 in macrophage activity and CD8 + T cell immune exhaustion. PCR validation further confirmed the differential expression of these genes. These findings offer new insights into the molecular mechanisms of CRC and provide a potential theoretical basis for targeting VSIG4 in immunotherapy.

Identification of PANoptosis related biomarkers to predict hepatic ischemia‒reperfusion injury after liver transplantation

Hepatic ischemia-reperfusion injury (HIRI) is a major complication following liver transplantation. Bioinformatic analysis was performed to elucidate the PANoptosis-related molecular mechanisms underlying HIRI. Comprehensive analysis of bulk and single-cell RNA sequencing data from human liver tissue before and after HIRI was performed. Differential expression analysis, weighted gene coexpression analysis, and protein interaction network analysis were used to identify candidate biomarkers. Multiple machine learning methods were utilized to screen for core biomarkers and construct a diagnostic predictive model. Functional and interaction analyses of the genes were also performed. Cellular clustering and annotation, pseudotemporal trajectory, and intercellular communication analyses of HIRI were conducted. Six PANoptosis-associated genes (CEBPB, HSPA1A, HSPA1B, IRF1, SERPINE1, and TNFAIP3) were identified as HIRI-related biomarkers. These biomarkers are regulated by NF-κB and miRNA-155. A nomogram for HIRI prediction based on these biomarkers was constructed and validated. In addition, the heterogeneity and dynamic changes in macrophage subpopulations during HIRI were revealed, highlighting the roles of Kupffer cells and monocyte-derived macrophages in modulating the hepatic microenvironment. The MIF and VISFATIN signaling pathways play important roles in the interaction between macrophages and other cells. These findings enhance our understanding of the mechanisms of PANoptosis in HIRI and provide a new basis and potential targets for prevention and treatment strategies for HIRI.

Interpretable Machine Learning reveals the Role of PANoptosis in the Diagnosis and Subtyping of NAFLD.

Interpretable Machine Learning reveals the Role of PANoptosis in the Diagnosis and Subtyping of NAFLD.

A novel perspective of ATR-FTIR spectroscopy combined with multiple machine learning methods for postmortem interval (PMI) human skin

A novel perspective of ATR-FTIR spectroscopy combined with multiple machine learning methods for postmortem interval (PMI) human skin

A Machine Learning Model for Predicting Breast Cancer Recurrence and Supporting Personalized Treatment Decisions Through Comprehensive Feature Selection and Explainable Ensemble Learning.

This study investigates the efficiency of a machine learning model integrating least absolute shrinkage and selection operator (LASSO) feature selection with ensemble learning in predicting recurrence risk and supporting personalized treatment decisions in breast cancer patients. Clinical data from 1,131 breast cancer patients (1,056 nonrecurrent and 75 recurrent) were collected from Kaohsiung Medical University Hospital's electronic health record system. After preprocessing and standardization, LASSO was applied for feature selection. An ensemble learning model was developed based on multiple machine learning algorithms, with SHAP (Shapley additive explanations) used for interpretability. The ensemble model achieved an AUC of 0.817, outperforming the best single model (AUC 0.711), demonstrating improved predictive accuracy and stability. LASSO identified six key predictors: regional lymph node positivity, ER status, Ki-67, lymphovascular invasion, tumor size, and age at diagnosis. SHAP analysis enhanced transparency by quantifying the contribution of each feature to recurrence risk, improving clinical understanding. This LASSO-enhanced ensemble model significantly improves the accuracy and interpretability of breast cancer recurrence prediction. By identifying individualized recurrence risks through SHAP analysis, the model supports more precise, data-driven clinical decision-making. These findings demonstrate its potential as a clinical decision support tool for guiding personalized treatment strategies, contributing to more effective breast cancer management.