Search Results (380)

Artificial intelligence (AI) techniques—ranging from hybrid mechanistic–machine learning (ML) ensembles to gradient-boosted decision trees, support-vector machines, and deep neural networks—are transforming the management of seasonal influenza, respiratory syncytial virus (RSV), human immunodeficiency virus (HIV), and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Symptom-based triage models using eXtreme Gradient Boosting (XGBoost) and Random Forests, as well as imaging classifiers built on convolutional neural networks (CNNs), have improved diagnostic accuracy across respiratory infections. Transformer-based architectures and social media surveillance pipelines have enabled real-time monitoring of COVID-19. In HIV research, support-vector machines (SVMs), logistic regression, and deep neural network (DNN) frameworks advance viral-protein classification and drug-resistance mapping, accelerating antiviral and vaccine discovery. Despite these successes, persistent challenges remain—data heterogeneity, limited model interpretability, hallucinations in large language models (LLMs), and infrastructure gaps in low-resource settings. We recommend standardized open-access data pipelines and integration of explainable-AI methodologies to ensure safe, equitable deployment of AI-driven interventions in future viral-outbreak responses. Full article

The rapid evolution of SARS-CoV-2 has underscored the need for a detailed understanding of antibody binding mechanisms to combat immune evasion by emerging variants. In this study, we investigated the interactions between Class I neutralizing antibodies—BD55-1205, BD-604, OMI-42, P5S-1H1, and P5S-2B10—and the receptor-binding domain (RBD) of the SARS-CoV-2 spike protein using multiscale modeling, which combined molecular simulations with the ensemble-based mutational scanning of the binding interfaces and binding free energy computations. A central theme emerging from this work is that the unique binding strength and resilience to immune escape of the BD55-1205 antibody are determined by leveraging a broad epitope footprint and distributed hotspot architecture, additionally supported by backbone-mediated specific interactions, which are less sensitive to amino acid substitutions and together enable exceptional tolerance to mutational escape. In contrast, BD-604 and OMI-42 exhibit localized binding modes with strong dependence on side-chain interactions, rendering them particularly vulnerable to escape mutations at K417N, L455M, F456L and A475V. Similarly, P5S-1H1 and P5S-2B10 display intermediate behavior—effective in some contexts but increasingly susceptible to antigenic drift due to narrower epitope coverage and concentrated hotspots. Our computational predictions show strong agreement with experimental deep mutational scanning data, validating the accuracy of the models and reinforcing the value of binding hotspot mapping in predicting antibody vulnerability. This work highlights that neutralization breadth and durability are not solely dictated by epitope location, but also by how binding energy is distributed across the interface. The results provide atomistic insight into mechanisms driving resilience to immune escape for broadly neutralizing antibodies targeting the ACE2 binding interface—which stems from cumulative effects of structural diversity in binding contacts, redundancy in interaction patterns and reduced vulnerability to mutation-prone positions. Full article

Human Interleukin-6 (hIL-6) is a pro inflammatory cytokine that binds to its receptor, IL-6Rα followed by binding to gp130 and subsequent dimerization to form a hexamer signaling complex. As a critical inflammation mediator, hIL-6 is associated with a diverse range of diseases and monoclonal antibodies in clinical use that either target IL-6Rα or hIL-6 to inhibit signaling. Here, we perform high-throughput structure-based computational screening using ensemble docking for small-molecule antagonists for which the target conformations were taken from 600 ns long molecular dynamics simulations of the apo protein. Prior knowledge of the contact sites from binary complex studies and experimental work was incorporated into the docking studies. The top 20 scoring ligands from the in silico studies after post analysis were subjected to in vitro functional assays. Among these compounds, the ligand with the second-highest calculated binding affinity experimentally showed an ~84% inhibitory effect on IL6-induced STAT3 reporter activity at 10 μM concentration. This finding may pave the way for designing small-molecule inhibitors of hIL-6 of therapeutic significance. Full article

Skin cancer, particularly melanoma, remains a major public health concern due to its high mortality rate. Current treatment options, including chemotherapy with dacarbazine and doxorubicin, have shown limited efficacy, achieving only a 20% objective response rate over six months, along with severe side effects such as cardiotoxicity. Given these limitations, there is a growing interest in herbal medicine as a source of novel anticancer compounds. Bambusa stenostachya, a bamboo species native to Taiwan, was investigated for its potential anti-melanoma properties using network pharmacology and molecular docking. LC-MS analysis identified seven bioactive compounds, including quinic acid and isovitexin, which satisfied Lipinski’s drug-likeness criteria. Among the seven bioactive compounds identified, five belong to the flavonoid family, while two are classified as phenolic compounds that modulate signaling pathways related to cancer and exhibit antioxidant activity, respectively. Through pathway enrichment analysis, four key melanoma-associated genes (PIM1, MEK1, CDK2, and PDK1) were identified as potential therapeutic targets. Ensemble docking results demonstrated that naringin-7-rhamnoglucoside exhibited the highest binding affinity (−6.30 kcal/mol) with phosphoinositide-dependent kinase-1, surpassing the affinities of standard chemotherapeutic agents. Additionally, the average docking scores for naringin-7-rhamnoglucoside and the remaining three proteins were as follows: PIM1 (−5.92), MEK1 (−6.07), and CDK2 (−5.26). These findings suggest that the bioactive compounds in B. stenostachya may play a crucial role in inhibiting melanoma progression by modulating metabolic and signaling pathways. Further in vitro and in vivo studies are necessary to validate these computational findings and explore the potential of B. stenostachya as a complementary therapeutic agent for melanoma. Full article

Pine nuts hold significant economic value due to their rich plant protein and healthy fats, yet precise variety classification has long been hindered by limitations of traditional techniques such as chemical analysis and machine vision. This study proposes a novel near-infrared (NIR) spectral feature selection algorithm, termed the improved binary equilibrium optimizer with selection probability (IBiEO-SP), which incorporates a dynamic probability adjustment mechanism to achieve efficient feature dimensionality reduction. Experimental validation on a dataset comprising seven pine nut varieties demonstrated that, compared to particle swarm optimization (PSO) and the genetic algorithm (GA), the IBiEO-SP algorithm improved average classification accuracy by 5.7% (p < 0.01, Student’s t-test) under four spectral preprocessing methods (MSC, SNV, SG1, and SG2). Remarkably, only 2–3 features were required to achieve optimal performance (MSC + random forest: 99.05% accuracy, 100% F1/precision; SNV + KNN: 97.14% accuracy, 100% F1/precision). Furthermore, a multimodal data synergy strategy integrating NIR spectroscopy with morphological features was proposed, and a classification model was constructed using a soft voting ensemble. The final classification accuracy reached 99.95%, representing a 2.9% improvement over single-spectral-mode analysis. The results indicate that the IBiEO-SP algorithm effectively balances feature discriminative power and model generalization needs, overcoming the contradiction between high-dimensional data redundancy and low-dimensional information loss. This work provides a high-precision, low-complexity solution for rapid quality detection of pine nuts, with broad implications for agricultural product inspection and food safety. Full article

Background: Pediatric acute appendicitis (AAP) is a common cause of abdominal pain in children, yet accurate classification into negative, uncomplicated, and complicated forms remains clinically challenging. Misclassification may lead to unnecessary surgeries or delayed treatment. This study aims to evaluate and compare the diagnostic accuracy of five machine learning models (AdaBoost, XGBoost, Stochastic Gradient Boosting, Bagged CART, and Random Forest) for classifying pediatric AAP subtypes. Methods: In this retrospective observational study, a dataset of 590 pediatric patients was analyzed. Demographic information and laboratory parameters—including C-reactive protein (CRP), white blood cell (WBC) count, neutrophils, lymphocytes, and appendiceal diameter—were included as features. The cohort consisted of negative (19.8%), uncomplicated (49.2%), and complicated (31.0%) AAP cases. Five ensemble machine learning models (AdaBoost, XGBoost, Stochastic Gradient Boosting, Bagged CART, and Random Forest) were trained on 80% of the dataset and tested on the remaining 20%. Model performance was evaluated using accuracy, sensitivity, specificity, and F1 score, with cross-validation employed to ensure result stability. Results: Random Forest demonstrated the highest overall accuracy (90.7%), sensitivity (100.0%), and specificity (61.5%) for distinguishing negative and uncomplicated AAP cases. Meanwhile, XGBoost outperformed other models in identifying complicated AAP cases, achieving an accuracy of 97.3%, sensitivity of 100.0%, and specificity of 78.3%. The most influential biomarkers were neutrophil count, appendiceal diameter, and WBC levels, highlighting their predictive value in AAP classification. Conclusions: ML models, particularly Random Forest and XGBoost, exhibit strong potential in aiding pediatric AAP diagnosis. Their ability to accurately classify AAP subtypes suggests that ML-based decision support tools can complement clinical judgment, improving diagnostic precision and patient outcomes. Future research should focus on multi-center validation, integrating imaging data, and enhancing model interpretability for broader clinical adoption. Full article

(1) RNA-binding proteins (RBPs) play a crucial role in regulating gene expression in plants, affecting growth, development, and stress responses. Accurate prediction of plant-specific RBPs is vital for understanding gene regulation and enhancing genetic improvement. (2) Methods: We propose an ensemble learning method that integrates shallow and deep learning. It integrates prediction results from SVM, LR, LDA, and LightGBM into an enhanced TextCNN, using K-Peptide Composition (KPC) encoding (k = 1, 2) to form a 420-dimensional feature vector, extended to 424 dimensions by including those four prediction outputs. Redundancy is minimized using a Pearson correlation threshold of 0.80. (3) Results: On the benchmark dataset of 4992 sequences, our method achieved an ACC of 97.20% and 97.06% under 5-fold and 10-fold cross-validation, respectively. On an independent dataset of 1086 sequences, our method attained an ACC of 99.72%, an ${\mathrm{F}1}_{\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}}$ of 99.72%, an MCC of 99.45%, an SN of 99.63%, and an SP of 99.82%, outperforming RBPLight by 12.98 percentage points in ACC and the original TextCNN by 25.23 percentage points. (4) Conclusions: These results highlight our method’s superior accuracy and efficiency over PSSM-based approaches, enabling large-scale plant RBP prediction. Full article

Metritis is one of the most common postpartum diseases in dairy cows, associated with impaired reproductive performance and substantial economic losses. In this study, we investigated the potential of machine learning (ML) techniques applied to physiological, behavioural, and milk quality parameters for the early detection of metritis in dairy cows during the postpartum period. A total of 2707 daily observations were collected from 94 cows in early lactation, of which 11 cows (275 records) were diagnosed with metritis. The dataset included daily measurements of body weight, rumination time, milk yield, milk composition (fat, protein, lactose), somatic cell count (SCC), and feed intake. Five classification models—partial least squares discriminant analysis (PLS-DA), random forest (RF), support vector machine (SVM), neural network (NN), and an Ensemble model—were developed using standardised features and stratified 80/20 training/test splits. To address class imbalance, model loss functions were adjusted using class weights. Models were evaluated based on accuracy, sensitivity, specificity, positive and negative predictive values (PPV, NPV), area under the receiver operating characteristic (ROC) area under the curve (AUC), and Matthews correlation coefficient (MCC). The NN model demonstrated the highest overall performance (accuracy = 96.1%, AUC = 96.3%, MCC = 0.79), indicating strong capability in distinguishing both healthy and diseased animals. The SVM achieved the highest sensitivity (90.9%), while RF and Ensemble models showed high specificity (>98%) and PPV. This study provides novel evidence that ML methods can effectively detect metritis using routinely collected, non-invasive on-farm data. Our findings support the integration of neural and Ensemble learning models into automated health monitoring systems to enable earlier disease detection and improved animal welfare. Although external validation was not performed, internal cross-validation demonstrated consistent performance across models, suggesting suitability for application in multi-farm settings. To the best of our knowledge, this is among the first studies to apply ML for early metritis detection based exclusively only automated herd data. Full article

KRAS is a pivotal oncoprotein that regulates cell proliferation and survival through interactions with downstream effectors such as RAF1. Despite significant advances in understanding KRAS biology, the structural and dynamic mechanisms of KRAS allostery remain poorly understood. In this study, we employ microsecond molecular dynamics simulations, mutational scanning, and binding free energy calculations together with dynamic network modeling to dissect how engineered DARPin proteins K27, K55, K13, and K19 engage KRAS through diverse molecular mechanisms ranging from effector mimicry to conformational restriction and allosteric modulation. Mutational scanning across all four DARPin systems identifies a core set of evolutionarily constrained residues that function as universal hotspots in KRAS recognition. KRAS residues I36, Y40, M67, and H95 consistently emerge as critical contributors to binding stability. Binding free energy computations show that, despite similar binding modes, K27 relies heavily on electrostatic contributions from major binding hotspots while K55 exploits a dense hydrophobic cluster enhancing its effector-mimetic signature. The allosteric binders K13 and K19, by contrast, stabilize a KRAS-specific pocket in the α3–loop–α4 motif, introducing new hinges and bottlenecks that rewire the communication architecture of KRAS without full immobilization. Network-based analysis reveals a strikingly consistent theme: despite their distinct mechanisms of recognition, all systems engage a unifying allosteric architecture that spans multiple functional motifs. This architecture is not only preserved across complexes but also mirrors the intrinsic communication framework of KRAS itself, where specific residues function as central hubs transmitting conformational changes across the protein. By integrating dynamic profiling, energetic mapping, and network modeling, our study provides a multi-scale mechanistic roadmap for targeting KRAS, revealing how engineered proteins can exploit both conserved motifs and isoform-specific features to enable precision modulation of KRAS signaling in oncogenic contexts. Full article

The elucidation of protein dynamics, especially in the context of intrinsically disordered proteins, is challenging and requires cooperation between experimental studies and computational analysis. Molecular dynamics simulations are an essential investigation tool but often struggle to accurately quantify the conformational preferences of flexible proteins. To create a quantitatively validated conformational ensemble, such simulations may be refined with experimental data using Bayesian and maximum entropy methods. In this study, we present a method to optimize a conformational ensemble using Bayes’ theorem in connection with a methodology derived from Umbrella Sampling. The resulting method, called the Umbrella Refinement of Ensembles (URE), reduces the number of parameters to be optimized in comparison to the classical Bayesian Ensemble Refinement and remains methodologically suitable for use with the forward formulated Kullback–Leibler divergence. The method is validated using two established systems, an alanine–alanine zwitterion and the chignolin peptide, using nuclear magnetic resonance data from the literature. Full article

This paper presents a novel approach for screening women in their first trimester of pregnancy to identify those at high risk of having a child with Down syndrome (DS), using machine learning algorithms. Various machine learning models, including statistical, linear, and ensemble models, were trained using a pseudo-anonymized dataset of 90,532 screening patients. This dataset, containing less than 1% positive cases, was obtained from Cruces University Hospital, a public health hospital (Osakidetza) in Baracaldo, Basque Country, Spain. The models incorporate a set of input variables, including demographic variables such as maternal age, weight, ethnicity, smoking status, and diabetes status, as well as laboratory variables like nuchal translucency (NT), pregnancy-associated plasma protein-A (PAPP-A), and beta-human chorionic gonadotropin hormone (B-HCG) levels. The trained classification algorithms achieved ROC-AUC values between 0.970 and 0.982, with sensitivity and specificity of 0.94. The results indicate that machine learning techniques can effectively predict Down syndrome risk in first-trimester screening programs. Full article

Accurately predicting protein–ligand binding affinity is an important step in the drug discovery process. Deep learning (DL) methods have improved binding affinity prediction by using diverse categories of molecular data. However, many models rely heavily on interaction and sequence data, which impedes proper learning and limits performance in de novo applications. To address these limitations, we developed a novel graph neural network model, called StructureNet (structure-based graph neural network), to predict protein–ligand binding affinity. StructureNet improves existing DL methods by focusing entirely on structural descriptors to mitigate data memorization issues introduced by sequence and interaction data. StructureNet represents the protein and ligand structures as graphs, which are processed using a GNN-based ensemble deep learning model. StructureNet achieved a PCC of 0.68 and an AUC of 0.75 on the PDBBind v.2020 Refined Set, outperforming similar structure-based models. External validation on the DUDE-Z dataset showed that StructureNet can effectively distinguish between active and decoy ligands. Further testing on a small subset of well-known drugs indicates that StructureNet has high potential for rapid virtual screening applications. We also hybridized StructureNet with interaction- and sequence-based models to investigate their impact on testing accuracy and found minimal difference (0.01 PCC) between merged models and StructureNet as a standalone model. An ablation study found that geometric descriptors were the key drivers of model performance, with their removal leading to a PCC decrease of over 15.7%. Lastly, we tested StructureNet on ensembles of binding complex conformers generated using molecular dynamics (MD) simulations and found that incorporating multiple conformations of the same complex often improves model accuracy by capturing binding site flexibility. Overall, the results show that structural data alone are sufficient for binding affinity predictions and can address pattern recognition challenges introduced by sequence and interaction features. Additionally, structural representations of protein–ligand complexes can be considerably improved using geometric and topological descriptors. We made StructureNet GUI interface freely available online. Full article

RNA-dependent RNA polymerase (RdRP) represents a critical target for antiviral drug development. We developed a multi-model machine learning framework combining five traditional algorithms (ExtraTreesClassifier, RandomForestClassifier, LGBMClassifier, BernoulliNB, and BaggingClassifier) with a CNN deep learning model to identify potential RdRP inhibitors among FDA-approved drugs. Using the PubChem dataset AID 588519, our ensemble models achieved the highest performance with accuracy, ROC-AUC, and F1 scores higher than 0.70, while the CNN model demonstrated complementary predictive value with a specificity of 0.77 on external validation. Molecular docking studies with the norovirus RdRP (PDB: 4NRT) identified raloxifene as a promising candidate, with a binding affinity (−8.8 kcal/mol) comparable to the positive control (−9.2 kcal/mol). The molecular dynamics simulation confirmed stable binding with RMSD values of 0.12–0.15 nm for the protein–ligand complex and consistent hydrogen bonding patterns. Our findings suggest that raloxifene may possess RdRP inhibitory activity, providing a foundation for its experimental validation as a potential broad-spectrum antiviral agent. Full article

Background: Breast cancer (BC) is the most common cancer among women worldwide; therefore, the efforts of many scientists are aimed at finding effective biomarkers for this disease. It is known that exosomes are nanosized extracellular vesicles (EVs) that are released from various cell types, including cancer cells. Exosomes are directly involved in governing the physiological and pathological processes of an organism through the horizontal transfer of functional molecules (proteins, microRNA, etc.) from producing to receiving cells. Since the diagnosis and treatment of BC have been improved substantially with exosomes, in this study, we isolated breast carcinoma cell-derived exosomes, primary endotheliocyte-derived exosomes, and blood exosomes from BC patients (BCPs) in the first stage of disease and investigated their proteomic profiles. Methods: Exosomes were isolated from the samples by ultrafiltration and ultracentrifugation, followed by mass spectrometric and bioinformatics analyses of the data. The exosomal nature of vesicles was verified using transmission electron microscopy and flow cytometry. Results: Exosome proteins secreted by MCF-7 and BT-474 cells were found to form two clusters, one of which enhanced the malignant potential of cancer cells, while the other coincided with a cluster of HUVEC-derived exosome proteins. Despite the different ensembles of proteins in exosomes from the MCF-7 and BT-474 lines, the relevant portions of these proteins are involved in similar biological pathways. Comparison analysis revealed that more BC-associated proteins were found in the exosomal fraction of blood from BCPs than in the exosomal fraction of conditioned medium from cells mimicking the corresponding cancer subtype (89% and 81% for luminal A BC and MCF-7 cells and 86% and 80% for triple-positive BC and BT-474 cells, respectively). Conclusions: Tumor-associated proteins should be sought not in exosomes secreted by cell lines but in the composition of blood exosomes from cancer patients, while the contribution of endotheliocyte exosomes to the total pool of blood exosomes can be neglected. Full article

Near-UV circular dichroism (CD) spectroscopy is a widely used method that provides, among others, information about the tertiary structure of biomolecular systems such as proteins, RNA, or DNA. Experimental near-UV CD spectra of proteins reflect the CD signals averaged over the many conformations that these systems can adopt. Theoretical approaches have been developed to predict such spectroscopic properties and link modeled conformations of complex biosystems to easily accessible experimental data, without having the resort to costly structural biology techniques. However, these predictions are mostly generated on the basis of a single experimental structure, missing the dynamic information reflecting the protein conformational variability. Here, we describe a complete reformulation of the theoretical foundations behind the prediction of CD spectra. We propose a QM/MM-based automated pipeline that generates an average near-UV CD spectrum from a given MD ensemble in a fast manner based on these theoretical considerations and further test it on protein systems. This pipeline has been implemented in an open-source program called DichroProt. Full article