Chinese Journal of Chemical Engineering, Год журнала: 2025, Номер unknown
Опубликована: Июнь 1, 2025
Язык: Английский
Cell Reports Physical Science, Год журнала: 2025, Номер unknown, С. 102516 - 102516
Опубликована: Апрель 1, 2025
Язык: Английский
Процитировано
1
Journal of Chemical Information and Modeling, Год журнала: 2025, Номер unknown
Опубликована: Март 17, 2025
Anesthetics are crucial in surgical procedures and therapeutic interventions, but they come with side effects varying levels of effectiveness, calling for novel anesthetic agents that offer more precise controllable effects. Targeting Gamma-aminobutyric acid (GABA) receptors, the primary inhibitory receptors central nervous system, could enhance their action, potentially reducing while improving potency anesthetics. In this study, we introduce a proteomic learning GABA receptor-mediated anesthesia based on 24 receptor subtypes by considering over 4000 proteins protein-protein interaction (PPI) networks 1.5 millions known binding compounds. We develop corresponding drug-target network to identify potential lead compounds design. To ensure robust predictions, curated dataset comprising 136 targets from pool 980 within PPI networks. employed three machine algorithms, integrating advanced natural language processing (NLP) models such as pretrained transformer autoencoder embeddings. Through comprehensive screening process, evaluated repurposing 180,000 drug candidates targeting GABRA5 receptor. Additionally, assessed ADMET (absorption, distribution, metabolism, excretion, toxicity) properties these those near-optimal characteristics. This approach also involved optimizing structures existing Our work presents an innovative strategy development new drugs, optimization use, deeper understanding anesthesia-related
Язык: Английский
Процитировано
0
European Journal of Medicinal Chemistry Reports, Год журнала: 2025, Номер unknown, С. 100249 - 100249
Опубликована: Янв. 1, 2025
Язык: Английский
Процитировано
0
Journal of the American Medical Informatics Association, Год журнала: 2025, Номер unknown
Опубликована: Фев. 25, 2025
Abstract Objectives The vast and complex nature of human genomic sequencing data presents challenges for effective analysis. This review aims to investigate the application natural language processing (NLP) techniques, particularly large models (LLMs) transformer architectures, in deciphering codes, focusing on tokenization, models, regulatory annotation prediction. goal this is assess model accessibility most recent literature, gaining a better understanding existing capabilities constraints these tools data. Materials Methods Following Preferred Reporting Items Systematic Reviews Meta-Analyses (PRISMA) guidelines, our scoping was conducted across PubMed, Medline, Scopus, Web Science, Embase, ACM Digital Library. Studies were included if they focused NLP methodologies applied analysis, without restrictions publication date or article type. Results A total 26 studies published between 2021 April 2024 selected review. highlights that tokenization enhance data, with applications predicting annotations like transcription-factor binding sites chromatin accessibility. Discussion LLMs interpretation promising field can help streamline large-scale while also providing its structures. It has potential drive advancements personalized medicine by offering more efficient scalable solutions Further research needed discuss overcome current limitations, enhancing transparency applicability. Conclusion growing role NLP, LLMs, While improve prediction, remain interpretability. refine their genomics.
Язык: Английский
Процитировано
0
Discover Oncology, Год журнала: 2025, Номер 16(1)
Опубликована: Март 17, 2025
This study proposes an advanced machine learning (ML) framework for breast cancer diagnostics by integrating transcriptomic profiling with optimized feature selection and classification techniques. A dataset of 1759 samples (987 patients, 772 healthy controls) was analyzed using Recursive Feature Elimination, Boruta, ElasticNet selection. Dimensionality reduction techniques, including Non-Negative Matrix Factorization (NMF), Autoencoders, transformer-based embeddings (BioBERT, DNABERT), were applied to enhance model interpretability. Classifiers such as XGBoost, LightGBM, ensemble voting, Multi-Layer Perceptron, Stacking trained grid search cross-validation. Model evaluation conducted accuracy, AUC, MCC, Kappa Score, ROC, PR curves, external validation performed on independent 175 samples. XGBoost LightGBM achieved the highest test accuracies (0.91 0.90) AUC values (up 0.92), particularly NMF BioBERT. The Voting method exhibited best accuracy (0.92), confirming its robustness. Transformer-based techniques significantly improved performance compared conventional approaches like PCA Decision Trees. proposed ML enhances diagnostic interpretability, demonstrating strong generalizability dataset. These findings highlight potential precision oncology personalized diagnostics.
Язык: Английский
Процитировано
0
European Journal of Medicinal Chemistry, Год журнала: 2025, Номер 291, С. 117615 - 117615
Опубликована: Апрель 10, 2025
Transformer-based chemical language models (CLMs) were derived to generate structurally and topologically diverse embeddings of core structure fragments, substituents, or core/substituent combinations in chemically proper compounds, representing a design task that is difficult address using conventional generation methods. To this end, CLM variants challenged learn different fragment-to-compound mappings the absence structural rules any other fragment linking synthetic information. The resulting alternative found have high syntactic fidelity, but displayed notable differences their ability valid candidate compounds containing test with clear preference for model variant processing combinations. However, majority generated all distinct from training data novel. In addition, CLMs exhibited diversification capacity often structures new topologies not encountered during training. Furthermore, produced large numbers close analogues known bioactive covering target space, thus indicating relevance newly candidates pharmaceutical research. As part our study, methodology are made publicly available.
Язык: Английский
Процитировано
0
Macromolecular Bioscience, Год журнала: 2025, Номер unknown
Опубликована: Апрель 22, 2025
Abstract Computer‐aided protein design (CAPD) has become a transformative field, harnessing advances in computational power and deep learning to deepen the understanding of structure, function, design. This review provides comprehensive overview CAPD techniques, with focus on their application protein‐based therapeutics such as monoclonal antibodies, drugs, antigens, polymers. starts key methods, particularly those integrating learning‐based predictions generative models. These approaches have significantly enhanced drug properties, including binding affinity, specificity, reduction immunogenicity. also covers CAPD's role optimizing vaccine antigen design, improving stability, customizing polymers for delivery applications. Despite considerable progress, faces challenges model overfitting, limited data rare families, need efficient experimental validation. Nevertheless, ongoing advancements coupled interdisciplinary collaborations, are poised overcome these obstacles, advancing engineering therapeutic development. In conclusion, this highlights future potential transform development, personalized medicine, biotechnology.
Язык: Английский
Процитировано
0
Journal of Applied Toxicology, Год журнала: 2025, Номер unknown
Опубликована: Май 1, 2025
The prediction of chemical toxicity is crucial for applications in drug discovery, environmental safety, and regulatory assessments. This study aims to evaluate the performance advanced deep learning architectures, TabNet TabTransformer, comparison traditional machine methods, predicting compounds across 12 toxicological endpoints. dataset consisted 12,228 training 3057 test samples, each characterized by 801 molecular descriptors representing structural features. Traditional models, including XGBoost, CatBoost, SVM, a voting classifier, were paired with feature selection techniques such as principal component analysis (PCA), recursive elimination (RFE), mutual information (MI). Advanced trained directly on full set without dimensionality reduction. Model was assessed using accuracy, F1-score, AUC-ROC, AUPR, Matthews correlation coefficient (MCC), alongside SHAP interpret importance enhance model transparency under class imbalance conditions. Cross-validation evaluations ensured robust comparisons all models TabTransformer consistently outperformed classifiers, achieving AUC-ROC values up 96% endpoints SR.ARE SR.p53. showed highest complex labels, benefiting from self-attention mechanisms that captured intricate relationships, while achieved competitive outcomes an efficient, dynamic selection. In addition standard metrics, we reported AUPR MCC better imbalance, both maintaining high scores Although particularly performed well when combined selection-achieving 94% SR.p53-they lagged behind generalizability interaction modeling. further highlighted interpretability proposed architectures identifying influential VSAEstate6 MoRSEE8. highlights superiority ensuring through analysis. These offer promising alternative vitro vivo approaches, paving way cost-effective ethical
Язык: Английский
Процитировано
0
Frontiers in Pediatrics, Год журнала: 2025, Номер 13
Опубликована: Май 23, 2025
Background Computer vision (CV), a subset of artificial intelligence (AI), enables deep learning models to detect specific events within digital images or videos. Especially in medical imaging, AI/CV holds significant promise analyzing data from x-rays, CT scans, and MRIs. However, the application support surgery has progressed more slowly. This study presents development first image-based model classifying quality indicators laparoscopic Nissen fundoplication (LNF). Materials methods Six visible (VQIs) for were predefined as parameters build datasets including correct (360° fundoplication) incorrect configurations (incomplete, twisted wraps, too long (>four knots), loose, long, malpositioning (at/below gastroesophageal junction). In porcine model, multiple iterations each VQI performed. A total 57 video sequences processed, extracting 3,138 at 0.5-second intervals. These annotated corresponding their respective VQIs. The EfficientNet architecture, typical was employed train an ensemble image classifiers, well multi-class classifier, distinguish between wraps. Results demonstrated strong performance predicting VQIs fundoplication. individual classifiers achieved average F1-Score 0.9738 ± 0.1699 when adjusted optimal Equal Error Rate (EER) decision boundary. similar observed using classifier. results remained robust despite extensive augmentation. For 3/5 identical; detection incomplete loose LNFs showed slight decline predictive power. Conclusion experimental demonstrates that algorithm can effectively fundoplications. proof concept does not aim test clinical fundoplication, but provides evidence be trained classify various surgical configurations. future, this could developed into AI based real-time enhance outcome patient safety.
Язык: Английский
Процитировано
0
Chinese Journal of Chemical Engineering, Год журнала: 2025, Номер unknown
Опубликована: Июнь 1, 2025
Язык: Английский
Процитировано
0