IntroductionGastric Cancer (GC) is a major global health problem with high incidence and mortality, making it necessary to find new prognostic biomarkers. Glycoprotein Non-Metastatic Melanoma Protein B (GPNMB) has been linked to tumor growth in several cancers, but its role as a prognostic marker in GC remains unclear. MethodsTo explore GPNMB’s expression, immune microenvironment association, and prognostic value in GC, this study integrated bioinformatics (TCGA, GEO, GEPIA) and Immunohistochemistry (IHC) on 73 paired GC/adjacent tissues, with semi-quantitative scoring and statistical analyses (chi-square, Mann-Whitney U, log-rank, multivariate Cox regression). ResultsGPNMB was significantly overexpressed in GC tissues vs. adjacent tissues (IHC: 57 ± 53 vs. 38 ± 39, p = 0.018; consistent with TCGA/GEO). High GPNMB correlated with abnormal immune/stromal scores (F = 6.55, Pr(>F) = 0.000249), increased immunosuppressive cell infiltration, decreased effector T cells, advanced M stage (distant metastasis rate: 24.2% vs. 5.0%, p = 0.017), shorter overall survival (log-rank p = 0.018), and was an independent prognostic factor (HR = 1.52, 95% CI:1.03–2.24, p = 0.035; multi-dataset HR:1.446–1.7, p < 0.05). Females showed a higher proportion of GPNMB (36.4% vs. 17.5% males), but this difference was not significant (p = 0.068). DiscussionOur study confirms GPNMB’s overexpression in gastric cancer via integrated bioinformatics and IHC, linking it to immune microenvironment dysregulation, distant metastasis, and poor prognosis-reinforcing its role as an independent prognostic biomarker and potential immunotherapy target, while noting limitations of small single-center IHC cohort, lack of experimental validation for relevant signaling pathway involvement, and reliance on algorithmic immune infiltration analyses. ConclusionGPNMB is overexpressed in GC, drives progression (especially metastasis), and shortens survival by regulating the immune microenvironment to exacerbate immune escape. It serves as an independent GC prognostic marker for risk stratification and a potential target for combined immunotherapy.
IntroductionPaper money has served as a fomite and vector for centuries, potentially harboring pathogenic microorganisms due to contamination with nosocomial infections, such as biofilm-forming, multidrug-resistant bacteria. This study was conducted to examine environmental contamination of Iraqi current currency notes with bacteria, biofilm-forming genes, and antimicrobial-resistant profiles. MethodsIn this cross-sectional study, 32 Iraqi currency notes were obtained from six commercial sectors (October 2024–February 2025). Bacterial identification was performed using standard methods and the VITEK-2 automated system. Antimicrobial susceptibility testing was performed using the Kirby-Bauer disk diffusion method (CLSI 2024 criteria). PCR amplification was used to detect the icaA gene. Statistical analysis was performed using the chi-square and Fisher's exact tests (p-value < 0.05). ResultsAll 32 currency notes (100%) were positive for contamination, yielding a total of 50 bacterial isolates. Overall, Staphylococcus species were the most frequently isolated organism (n = 29, 58%), with S. epidermidis representing the most common isolate (n = 19, 38%). High rates of resistance were recorded for oxacillin (86.2%) and erythromycin (79.3%). In terms of multi-drug resistance, fifteen (51.7%) isolates were multi-drug-resistant (MDR). The icaA gene was present in 18 of 29 (62.1%) Staphylococcus isolates and was significantly associated with robust biofilm production (p-value < 0.001). DiscussionThe presence of multidrug resistance and the biofilm-associated gene IcaA among the isolates indicated an apparent public health risk. This reflects the highly active, community-wide circulation of antimicrobial-resistant and highly persistent pathogens that survive both antibiotics and host defenses during everyday interactions. ConclusionThe significant burden of these high-risk bacteria on Iraqi currency underscores the urgent need to raise hygiene levels at the community level and to develop effective control strategies to prevent money from serving as a vehicle for pathogen transmission.
Introduction Artificial intelligence (AI) has rapidly emerged as a transformative force in biomedical research, driving advances in data interpretation, diagnostics, and therapeutic strategies. The objective of this study was to conduct a large-scale bibliometric analysis of AI-related biomedical publications indexed in PubMed between 1950 and 2024, with the aim of identifying temporal trends, global research contributions, and thematic focus areas. Methods A retrospective bibliometric analysis was performed on 76,722 PubMed-indexed articles. Data were retrieved using a Python-based pipeline integrating National Center for Biotechnology Information (NCBI) Entrez Programming Utilities and Biopython. Metadata-including PubMed Identifier (PMID), title, abstract, authors, publication date, journal, Medical Subject Headings (MeSH) terms, Digital Object Identifier (DOI), and country of origin-were extracted, cleaned, and compiled into a structured dataset. Articles were analyzed for completeness, publication patterns, geographic distribution, journal outlets, and thematic focus using MeSH keywords. Results AI-related publications remained scarce until 2015, after which output expanded exponentially, reaching 20,135 articles in 2024. The United States contributed the largest share (32.5%, n = 24,958), followed by England (22.5%) and Switzerland (n = 13,936). Thematic analysis revealed machine learning (n = 16,281), deep learning (n = 12,297), and neural networks (n = 7,117) as dominant areas, while AI ethics appeared in only 110 publications. Metadata completeness varied, with notable gaps in MeSH indexing (41.6%), abstracts (9%), and Digital Object Identifiers (DOI) (2.6%). Research was widely disseminated across more than 6,000 journals, with Sensors , Scientific Reports , and Public Library of Science ONE ( PLOS ONE) as leading outlets. Discussion The findings highlight AI’s transition from a peripheral topic to a core pillar of biomedical science, with rapid growth driven by technological advances and global health demands. Despite widespread adoption across multiple disciplines, gaps remain in ethical engagement and equitable global representation. Metadata inconsistencies also pose challenges for systematic synthesis and bibliometric analyses. Conclusion AI has become a central component of biomedical research, characterized by exponential growth, interdisciplinary adoption, and global expansion. However, limited attention to AI ethics and persistent disparities in research representation underscore the need for targeted policy, funding, and governance strategies. Continued bibliometric monitoring is essential to ensure responsible and inclusive integration of AI into clinical and research practice.
Introduction/Objective Quantitative whole-body MRI relies on accurate delineation of multiple anatomical structures, yet manual labeling is slow and variable. We evaluate AISHANet, a deep learning model for multistructure 3D segmentation, and compare it with SegResNet, UNETR, and UNet using the same dataset split and evaluation protocol. The task covers 14 muscle groups and both lungs. Methods The dataset includes 100 whole-body DIXON T1 axial volumes acquired on a 3T Philips scanner (one volume per patient) with reference annotations produced by five expert radiologists. We used 80 volumes for training, 12 for validation, and 8 for testing. Performance was assessed with Dice Similarity Coefficient (DSC), directed Hausdorff distance, Sensitivity, ROC AUC, and F1-score. Metrics were computed per patient, macro-averaged across the 16 structures, and summarized as mean ± standard deviation (SD) across test patients. Results AISHANet obtained the highest overall scores, with a mean DSC 0.871 ± 0.017, directed Hausdorff distance in millimeters 24.11 ± 10.92, sensitivity 0.894 ± 0.047, ROC AUC 0.947 ± 0.024, and F1-score 0.871 ± 0.061. The best performance was observed in larger muscle groups (gluteus, thighs, calves), where DSC exceeded 0.88. Discussion While AISHANet consistently outperformed the baselines, performance decreased in anatomically challenging regions (abdomen and back), which are affected by lower contrast and thinner structures in axial views. Across models, we observed different failure modes: SegResNet tended to produce smoother masks, UNETR reduced isolated false positives, and UNet showed higher sensitivity to anatomical variability. Conclusion Under a controlled, single-protocol comparison, AISHANet provided the highest overall accuracy for multistructure whole-body MRI segmentation in this dataset. Remaining errors in low-contrast and anatomically complex regions motivate future work on improving robustness and validating performance across additional imaging settings, including other MRI scanner manufacturers and additional MRI sequence types.
Introduction The accurate detection of somatic Copy Number Variations (CNVs) is a challenging task in cancer genomics. This study addresses the significant variability in performance and the lack of consensus among computational tools for somatic CNV calling. Methods We conducted a comprehensive benchmark evaluation of four widely used tools - CNVkit, Sequenza, Facets, and ASCAT. Their performance was assessed in terms of recall, precision, reproducibility, and inter-tool concordance using an orthogonally validated real-world dataset derived from the HCC1395 cell line. Results Our analysis revealed considerable differences in tool performance. Facets and Sequenza showed the most balanced accuracy and the highest reproducibility. In contrast, we observed poor consensus among tools, particularly for amplifications, where pairwise concordance values were frequently below 0.6. CNVkit showed high sensitivity for deletions but exhibited critically low and unstable performance for amplifications. Discussion The results show that tool selection is a primary source of variability in CNV studies, which can significantly impact downstream biological interpretation. The high discordance rates, especially for amplifications, highlight the inherent limitations and the risk of false negatives when relying on a single algorithm. Conclusion We conclude that for reliable somatic CNV detection, tools such as Facets or Sequenza are required. We also recommend adopting a consensus-based approach to reduce error rates and improve the quality of findings from individual algorithms.
Infant mortality is a pivotal indicator of community health and socioeconomic conditions. Despite global advancements in healthcare and significant reductions in infant mortality rates, substantial disparities persist, particularly in underserved populations. This research aims to tackle these disparities by enhancing the predictive accuracy of public health interventions. We utilize advanced feature selection techniques to identify critical predictors of infant survival, thereby supporting the development of targeted and effective health policies and practices. We introduce an enhanced Binary Multi-Objective Cheetah Optimization algorithm (BMOCO), specifically designed for feature selection in extensive medical datasets. The suggested method focuses on optimizing eight S-shaped and V-shaped transfer functions to refine the conversion of continuous position vectors into binary form, ensuring precise feature selection and robust model performance. The BMOCO method demonstrates superior accuracy and effectiveness in feature selection compared to traditional evolutionary optimization algorithms such as MOGA (Multi-Objective Genetic Algorithm), MOALO (Multi-Objective Ant Lion Optimizer), NSGA-II (Non-dominated Sorting Genetic Algorithm II), and MOQBHHO (Multi-Objective Quadratic Binary Harris Hawk Optimization). Applied to U.S. infant birth data, our approach achieves an average classification accuracy of 99.54%. Critical factors impacting infant mortality identified include maternal literacy, prenatal care frequency, and pre-existing maternal conditions, such as diabetes, smoking during pregnancy, body mass index, infant birth weight, and breastfeeding practices. These findings indicate that the optimized BMOCO model provides interpretable and data-driven insights that align with established clinical evidence. The results underscore the effectiveness of advanced machine learning techniques in uncovering significant health predictors. The proposed BMOCO algorithm offers a robust and interpretable tool for health professionals to enhance predictive models, facilitating targeted interventions to reduce infant mortality rates and improve public health outcomes.
IntroductionColorectal cancer is a highly complex disease that continues to rise in prevalence, posing significant difficulties in its management and treatment outcomes. Reduced expression of the SLC4A4 gene has been identified as a critical factor in driving tumor development and poor clinical prognosis in CRC. This reduction disrupts several key cellular mechanisms, including cellular proliferation, programmed cell death, and metastasis, largely by altering pH regulation within the cells. Given its influence on tumor behavior and patient survival, SLC4A4 expression levels could serve as a reliable prognostic marker in colorectal cancer. MethodsThe CryoEM structure of SLC4A4, retrieved from the Protein Data Bank (PDB), was refined using SwissModel and subjected to virtual drug screening via the DrugRep web server. This screening employed a database comprising FDA-approved drugs. ResultsNilotinib emerged as the most potent inhibitor following further validation through redocking using CBDOCK2, ranking it highest among the top three results from DrugRep. The docking analysis yielded a score of −11.2 kcal/mol for Nilotinib. A 50-nanosecond molecular dynamics simulation further validated these findings, revealing that Nilotinib formed robust interactions with the SLC4A4 protein. The simulation yielded consistent results, with a root mean square deviation of around 0.30 nm, a root mean square fluctuation of near 0.5 nm, a compact radius of gyration between 3.8 and 4.0 nm, and stable solvent-accessible surface area profiles. These findings confirmed that the drug-protein complex maintained structural stability throughout the simulation. DiscussionThe computational findings suggest that Nilotinib binds effectively and stably to SLC4A4, indicating its potential to modulate the function of this protein in CRC. Its established safety profile as an FDA-approved drug supports the feasibility of drug repurposing. These results also reinforce the utility of integrating structure-based drug screening with molecular dynamics simulations to identify novel therapeutic agents for cancer. ConclusionNilotinib holds significant potential as a therapeutic agent for colorectal cancer and warrants further experimental investigation to validate its effectiveness.
IntroductionNeonatal infections remain a major threat in intensive care units (ICUs), often progressing rapidly and asymptomatically within the first hours of admission. Early detection is critical to improve outcomes, yet timely and reliable risk prediction remains a challenge. Method We developed an explainable machine learning retrospective observational cohort study for early prediction of neonatal infection using high-resolution data from the MIMIC-III database. Two-time windows, 30 and 120 minutes post-ICU admission, were analyzed. Physiological and hematological variables were aggregated, and missing data were imputed using Iterative Imputation. We used stratified five-fold cross-validation to test several classification models and feature importance and SHAP analysis to examine model interpretability. Result CatBoost demonstrated the best performance in the 30-minute window (F1-score = 0.76), and Gradient Boosting had the best performance in the 120-minute window (F1-score ≈ 0.80). Heart rate, white blood cell count, and temperature were important predictors because they showed both physiological stability and immune response. The 120-minute window made the model work better, indicating the importance of data availability for making accurate predictions. DiscussionCatBoost at the 30-minute window and Gradient Boosting at the 120-minute window could find a balance between speed and accuracy. Model-based imputation could handle high missing-value rates, but external validation is needed because the data came from only one center and the devices were different. ConclusionThis study proposes a two-stage decision-support system that adapts to data collected during early and later ICU admission periods. By combining accurate prediction with model interpretability, the framework may enable timely diagnosis and targeted interventions, ultimately reducing neonatal morbidity and mortality.
Nasopharyngeal carcinoma (NPC) is a malignant tumor with distinct molecular features, underscoring the need for reliable biomarkers to improve diagnosis, prognosis, and therapeutic strategies. We analyzed transcriptomic data from GEO datasets (GSE12452, GSE53819, and GSE102349) to identify diagnostic and prognostic biomarkers. Differential expression analysis was performed to detect potential markers, while survival analysis was conducted using Cox proportional hazards (Cox-PH) modeling and log-rank tests. Elastic Net regression was used to refine the gene signature. RNA-protein expression concordance was validated using the Cancer Cell Line Encyclopedia (CCLE) dataset. Differential expression analysis revealed 591 genes as potential diagnostic markers. Survival analysis identified 54 genes with dual diagnostic and prognostic relevance. Elastic Net regression refined this to an 11-gene signature, which stratified patients into high- and low-risk groups, significantly predicting progression-free survival (log-rank p = 0.0035). Five genes (BUB1B, GAS2L3, NFE2L3, OIP5, and PDGFRL) were identified as potential oncogenic drivers, while six (CD1D, CYP4B1, IL33, KLF2, NAPSB, and VILL) were implicated as tumor suppressors. Six genes (BUB1B, GAS2L3, IL33, OIP5, PDGFRL, and VILL) showed strong RNA-protein expression concordance in the CCLE dataset. This study reveals previously unreported cancer-associated genes (NAPSB, GAS2L3, NFE2L3, PDGFRL, CD1D, CYP4B1, KLF2) in NPC while validating established biomarkers (BUB1B, OIP5, IL33, VILL). Our findings expand NPC molecular characterization but require further clinical validation. This study presents a robust gene signature for NPC, offering valuable insights into tumor progression and providing a foundation for advancing diagnostic strategies, improving prognostic stratification, and developing targeted therapies.
Medical image fusion combines the data obtained from different imaging modalities such as Computed Tomography (CT), Positron Emission Tomography (PET), and Magnetic Resonance Imaging (MRI) into a single, informative image that aids clinicians in diagnosis and treatment planning. No single imaging modality can provide complete information on its own. This has led to the emergence of a research field focused on integrating data from multiple modalities to maximize information in a single, unified representation. CNN (Convolutional Neural Network) was applied to achieve robust and effective multi-modal image fusion. By delving into the principles and practical applications of this deep learning approach, the paper also provides a comparative analysis of CNN-based results with other conventional fusion techniques. CNN-based image fusion delivers far better results in terms of qualitative and quantitative analysis when compared with other conventional fusion methods. The paper also discusses future perspectives, emphasizing advancements in deep learning that could drive the evolution of CNN-based fusion and enhance its effectiveness in medical imaging. CNN-based multi-modal medical image fusion proves strong advantages over traditional methods in terms of feature preservation and adaptability. However, challenges such as data dependency, computational complexity, and generalization across modalities persist. Emerging trends like attention mechanisms and transformer models show promise in addressing these gaps. Future work should focus on improving interpretability and clinical applicability, ensuring that deep learning fusion methods can be reliably integrated into real-world diagnostic systems. Ultimately, this work underscores the potential of CNN-based fusion to improve patient outcomes and shape the future of medical imaging by advancing the understanding of multi-modal fusion.
Colorectal cancer is one of the global health threats and ranks among the deadliest diseases worldwide. The recognition and elimination of polyps at a primary, precancerous phase are, therefore, the key to preventing CRC. Most of these polyps differ in size and level of malignancy, thus failing to be detected by the commonly used screening methods. In this study, AI-driven tools were designed using deep learning models, such as VGG16, ResNet, and EfficientNet. They were validated using datasets obtained from JSS Hospital to enhance the accuracy of polyp recognition and decrease the probability of CRC, thereby improving patient outcomes. Fastai comes with an intuitive API, where most functions related to data preprocessing, building, and training a model are already built. Logs on training and validation losses, accuracies, and confidence scores of the performance metrics ensure the rigors of evaluation across multiple epochs of training. The results were impressive, with the deep learning models performing almost constantly at an accuracy of 99% in image classification. The robustness of the models is guaranteed because the balance between validation loss and training loss is attained. Hence, there is no overfitting or underfitting, guaranteeing reliable predictions. An interactive web platform was developed using Hugging Face with Gradio, and real-time predictions could be made by allowing users to upload images. The confusion matrix indicated that these models achieved nearly perfect classification performance. The VGG16 model performed with 99.48% accuracy, 100% precision, 97.95% recall, and an F1 score of 98.96%. The VGG19 model outperformed the former by a slight margin, displaying an accuracy of 99.69%, precision of 100%, recall of 98.76%, and an F1 score of 98.37%. ResNet18 and ResNet50 performed exceptionally well, achieving 99.79% accuracy, 100% precision, 99.17% recall, and 99.58% F1 score. The model with the best performance, with a solid score, was Efficient Net, scoring an accuracy of 99.9%. In the study, the effectiveness of deep CNN models was validated for polyp detection to aid in CRC prevention. These effectively and stably well-performing models, being provided to users by a very user-friendly platform, set a very good precedent for their broad application in the future. This milestone success and careful evaluation have led to an improvement in diagnostic processes and, therefore, health outcomes.
Introduction Substance-induced disorders (SID) have gained increasing attention in recent years; however, there is insufficient data to fully understand the long-term dynamics of these conditions. Materials and Methods In a retrospective observational cohort study, we analyzed a publicly available database (Treatment Episode Data Set – Admissions) between 1992 and 2022 to determine cases of SID and to assess the influence of various independent clinical factors. Results Three waves in the dynamics of SID were identified. The first wave was from 1992 to 2001. The number of SID was minimal, with a predominance of men with an average age of 30-34 years, using mainly heroin, methamphetamine, or phencyclidine. The second wave was from 2002 to 2014. The total number of SID increased by 6.5 times, the percentage of women increased, the average age shifted to 25-29 years, and the predominant substances were opiates, stimulants, and barbiturates. The third wave, from 2015 to 2022, was characterized by continued growth in SID cases, surpassing the number of cases related to alcohol-induced disorders, substance abuse, and substance dependence. Methamphetamine and novel psychoactive substances play the most important role in the third wave. Discussion In some cases, SID can transform into a chronic relapsing psychiatric disorder, which will require creating special units for the dispensary monitoring to detect the transformation, as well as to provide comprehensive combined care. Conclusion SID is an actual problem. The number of these cases is growing faster than the general population of drug abusers, depending on the general drug situation.
Background Leukemia, which is a blood cancer, is caused by the abnormal growth of white blood cells (WBCs), primarily found in the myeloid and fatty tissues of bone marrow. Microscopy is used by microbiologists and pathologists to examine the blood for the detection of leukemia. Blood cells are analyzed for morphological markers that aid in the detection and classification of leukemia. However, this method is time-consuming for malignancy prognosis and may be influenced by the clinical abilities and work experience of microbiologists. Aims and Objectives This research aimed to review and analyze various machine learning (ML) and deep learning (DL) approaches for the identification and categorization of different types of leukemia, particularly acute myeloid leukemia (AML) and chronic myeloid leukemia (CML), based on microscopic images of white blood cells (WBCs). It also aimed to evaluate the efficacy of various machine learning and deep learning classifiers for detecting acute and chronic myeloid leukemia and classifying different types of leukocytes. Methods In this study, a Support Vector Machine (SVM) classifier, representing traditional machine learning (ML) models, and a Convolutional Neural Network (CNN) classifier, based on deep learning (DL) algorithms, were employed to identify and classify myelogenous leukemia and different types of leukocytes. Results The algorithms utilizing the above-mentioned classifiers demonstrated significantly better performance metrics compared to other models. Conventional artificial intelligence (AI) approaches in medical image analysis have demonstrated effectiveness in accurately and reliably classifying biological images, such as microscopic blood cells, with greater precision and reliability. Conclusion CNNs achieved the highest accuracy, while SVMs excelled in precision among traditional methods. Combining both techniques also yielded great results. While accuracy is an important metric, it is not the only factor to consider. Overall, CNNs are more effective at detecting and classifying leukocytes and myelogenous leukaemia.
Introduction Immune checkpoint blockade targeting PD-1/PD-L1 has revolutionized cancer treatment; however, resistance remains a major clinical challenge. V-domain Immunoglobulin Suppressor of T cell Activation (VISTA), a B7 family member with high expression in tumor-infiltrating lymphocytes of ovarian cancer, has emerged as a promising alternative target for immunotherapeutic intervention. Materials and Methods We performed in silico screening of 9,397 DrugBank compounds against PD-L1 and VISTA using AutoDock Vina. The top candidates based on docking scores were assessed through 100 ns molecular dynamics simulations, and binding free energies were calculated via MM-PBSA. Results DB15637, DB12867, and DB06744 showed the strongest PD-L1 binding affinities (−7.33 to −7.87 kcal/mol) with average RMSD values of 8.89 Å, 8.94 Å, and 7.57 Å, respectively. DB00321 exhibited the highest affinity for VISTA (−7.31 kcal/mol) with an RMSD of 6.18 Å, maintaining stable interactions with key residues throughout the simulation. Discussion The identified compounds demonstrated favorable docking scores, dynamic stability, and binding free energies, suggesting their potential as PD-L1 and VISTA inhibitors. Dual checkpoint targeting could enhance antitumor immune responses in ovarian cancer, where both proteins contribute to immune evasion. Conclusion This in silico study identified promising candidates for PD-L1 and VISTA inhibition. These findings provide a computational basis for further experimental validation to confirm their therapeutic potential in the treatment of ovarian cancer.
Introduction Accurate classification of brain tumours using MRI scans is vital for early diagnosis and treatment. However, conventional deep learning models often require complete MRI sequences, which can prolong scan times and lead to patient discomfort or motion-related image degradation. Thus, enhancing diagnostic accuracy under faster scanning conditions is a critical research need. Therefore, this research aims to show how our proposed mechanism, namely Multiscale Parallel Feature Aggregation Network (MPFAN), accurately improves the diagnosis of classifying brain tumours while maintaining Magnetic Resonance Imaging (MRI) quality in fast MRI scanning. Methods This article proposed an MPFAN architecture that utilizes parallel branches to extract image features from different scales, using independent pathways with varied filters and movement steps. Feature combination blocks, feedback prevention mechanisms, and strict training constraints enhance system reliability. Results MPFAN achieved an accuracy of 97.4%, outperforming many existing brain tumour classification models. Performance improved steadily over training epochs, and optimizer comparisons showed Adam and Ada-Delta yielded the best results. Ablation studies confirmed that multiscale feature extraction, dropout regularization, and feature fusion significantly contribute to classification accuracy. Discussion The MPFAN model demonstrates superior performance due to its ability to effectively extract and integrate multiscale features. Its dual-branch architecture enables deeper contextual understanding, and its high accuracy validates its clinical potential. However, the model’s reliance on a single dataset and potential overfitting in later training epochs indicate the need for broader validation and optimization in real-world clinical environments. Conclusion The proposed MPFAN architecture enhances brain tumour classification by improving image processing efficiency and decision-making speed, making it a reliable and effective diagnostic tool.
IntroductionGenotype imputation improves the resolution of genetic data, but traditional methods are computationally intensive or compromise privacy. Deep learning alternatives are often too large for client-side deployment. In this study, FastImpute, a workflow for creating lightweight, reference-free imputation models, was developed that enables real-time, accessible genetic risk assessment on edge devices. MethodsUsing whole-genome sequencing data from 2,504 individuals in the 1000 Genomes Project, linear and logistic regression models were trained to impute single-nucleotide polymorphisms (SNPs) used in the breast cancer polygenic risk score PRS313_BC. Models used SNPs from commercial genotyping arrays, and performance was evaluated against sequencing data and benchmarked against Beagle. ResultsThe polygenic risk score (PRS) calculated with our linear model correlated strongly with the PRS from true sequencing data (R² = 0.86), significantly outperforming no imputation and minor allele frequency imputation (R² = 0.38). Our logistic model correctly identified 4 of 6 individuals in the top 1% of breast cancer risk, matching Beagle’s performance. DiscussionOur approach balances performance and efficiency, enabling deployment on personal devices and preserving user privacy through local data processing. This approach democratizes access to genetic risk assessment using direct-to-consumer data. However, this proof of concept requires validation across other genomic contexts before clinical use. ConclusionThe FastImpute pipeline demonstrates that lightweight models can enable real-time genetic risk assessment on edge devices.
Introduction The accurate identification of repeats and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPRs) has a profound impact on studying and understanding prokaryotic immune systems. Methods A model with feature extraction and scoring is trained, solved made as a tool. The Welch’s t-test is conducted. Results The length of the repeater, the copy number of the repeater, the starting position sequence of the repeater, and the repeater sequence as the features. The scoring formula The sequence with overlapping starting points and the highest score among the absolutely repeat sequences is selected as CRISPR, which is used as a tool to find CRISPR. Among 302 archaea, 199 obtained the same results as pilerCR using findCrispr; 86 obtained more CRISPRs than pilerCR. The Welch’s t-test shows that the count of Crisprs recognized by the findCrispr tool is significantly different, with t-stat > 0. Discussion The feature extraction is effective. The model performs well, and the tool findCrispr is inclined to find more repeaters. The algorithm is a specialized algorithm that is sensitive to finding CRISPR with a small number of duplicates and has low tolerance for long, scattered repeats. Conclusion Features are extracted, and a scoring system is established using the tool findCrisprrealized, which performs superiorly to pilerCR in the identification of CRISPRs with multiple calibration repeaters. The tool findCrispr is of great significance for studying the biological function and mechanism of CRISPR.
Introduction Skin disorders present a significant global public health concern, affecting millions of individuals and contributing to high morbidity rates. The application of artificial intelligence, particularly deep learning, has emerged as a promising avenue for the automated classification and detection of skin diseases, offering potential improvements in accuracy, speed, and cost-effectiveness in dermatological diagnostics. Methods This study aims to evaluate the performance of convolutional neural networks (CNNs) in classifying skin diseases. Four widely recognized architectures—ResNet50, InceptionV3, EfficientNetB0, and VGG16—were implemented and compared using a dataset comprising 1,159 dermoscopy images across eight disease categories. Models were trained using the Adam optimizer with a batch size of 32 over 20 epochs. Performance metrics were assessed and benchmarked against findings from existing literature. Results Among the evaluated models, EfficientNetB0 achieved the highest precision at 96.76%, followed by InceptionV3 and ResNet50 with accuracies around 93.5%. VGG16 demonstrated the lowest performance, achieving an accuracy of 84.32%. These results indicate that EfficientNetB0 offers superior feature extraction and generalization capabilities for dermatological image classification. Discussion The findings suggest that recent CNN architectures, particularly EfficientNetB0, can significantly enhance the accuracy of skin disease classification. These improvements may facilitate more effective and scalable diagnostic tools in dermatology. Limitations of this study include the relatively small dataset and limited class diversity, which may affect model generalizability. Conclusion EfficientNetB0 outperformed ResNet50, InceptionV3, and VGG16 in classifying skin diseases, highlighting its potential for clinical application. Future research should focus on expanding datasets, refining model architectures, and deploying automated skin disease screening systems in real-world healthcare settings.
Cancer is one of the leading causes of death worldwide, accounting for approximately 10 million deaths annually. The most prevalent type of cancer in women is breast cancer, and there are not any prospective vaccinations available for the treatment of this disease. This study aimed to identify potent substances derived from natural products, such as curcumin analogs, which are crucial for enriching drug discovery, particularly in the prevention of breast cancer. This study utilized twelve novel curcumin analogs, specifically dibenzylidene-cyclohexanones, to predict their biological activity against breast cancer. Based on Lipinski's rule of five, selected compounds were screened using ADMETlab 3.0 to assess their drug-likeness properties. Then, the selected compounds were subsequently subjected to pharmacophore modeling using LigandScout, followed by molecular docking studies with the human estrogen receptor alpha (ERα; PDB ID: 2IOG) using AutoDock. Curcumin and tamoxifen were included as reference compounds for comparison. Based on the research conducted, all of the curcumin analogs met the criteria of Lipinski’s rule of five, except compound 12. Compound 4 demonstrated the best potential as an anticancer agent against ERα, with a pharmacophore fit-score of 36.87 based on pharmacophore modeling and binding energy of -11.10 kcal, which was higher than tamoxifen (-10.45 kcal/mol) and curcumin (-9.18 kcal/mol) based on a molecular docking study. Exploring curcumin analogs as potential anti-breast cancer agents is crucial for drug discovery and development. This study suggests that curcumin analog compound 4 can act as a potent inhibitor against ERα. However, further in vitro studies are required to confirm the efficacy of this compound.
