Methods Mol Biol. 2025;2889:53-66. doi: 10.1007/978-1-0716-4322-8_5.

ABSTRACT

Understanding the spatial and temporal dynamics of gene expression is crucial for unraveling molecular mechanisms underlying various biological processes. While traditional methods have offered insights into gene expression patterns, they primarily focus on mature mRNA transcripts, lacking real-time visualization of newly synthesized or nascent transcription events. Recent advancements in monitoring nascent transcription in live cells provide valuable insights into transcriptional dynamics. However, such approaches are limited in mammalian embryos. Addressing this gap, we optimized a single molecule fluorescent in situ hybridization (smFISH) technique and coupled it with deep learning algorithms to automate detection of nascent transcription in mouse embryonic tissue samples. Our method enables precise quantification and comparison of nascent transcripts within tissue sections, offering reproducible results and potential applications in studying gene expression dynamics across various developmental stages.

PMID:39745605 | DOI:10.1007/978-1-0716-4322-8_5

Rheumatol Int. 2025 Jan 2;45(1):14. doi: 10.1007/s00296-024-05756-5.

ABSTRACT

Women are disproportionately affected by chronic autoimmune diseases (AD) like systemic lupus erythematosus (SLE), scleroderma, rheumatoid arthritis (RA), and Sjögren's syndrome. Traditional evaluations often underestimate the associated cardiovascular disease (CVD) and stroke risk in women having AD. Vitamin D deficiency increases susceptibility to these conditions. CVD risk prediction in AD can benefit from surrogate biomarker for coronary artery disease (CAD), such as carotid ultrasound. Due to non-linearity in the CVD risk stratification, we use artificial intelligence-based system using AD biomarkers and carotid ultrasound. Investigate the relationship between AD and CVD/stroke markers including autoantibody-influenced plaque load. Second, to study the surrogate biomarkers for the CAD and gather radiomics-based features such as carotid intima-media thickness (cIMT), and plaque area (PA). Third and final, explore the automated CVD/stroke risk identification using advanced machine learning (ML) and deep learning (DL) paradigms. Analysed biomarker data from women with AD, including carotid ultrasonography imaging, clinical parameters, autoantibody profiles, and vitamin D levels. Proposed artificial intelligence (AI) models to predict CVD/stroke risk accurately in AD for women. There is a strong association between AD duration and elevated cIMT/PA, with increased CVD risk linked to higher rheumatoid factor (RF) and anti-citrullinated peptide antibodies (ACPAs) levels. AI models outperformed conventional methods by integrating imaging data and disorder-specific factors. Interdisciplinary collaboration is crucial for managing CVD/stroke in women with chronic autoimmune diseases. AI-based assisted risk stratification methods may improve treatment decision-making and cardiovascular outcomes.

PMID:39745536 | DOI:10.1007/s00296-024-05756-5

Anal Chem. 2025 Jan 2. doi: 10.1021/acs.analchem.4c05632. Online ahead of print.

ABSTRACT

Metabolite identification from 1D 1H NMR spectra is a major challenge in NMR-based metabolomics. This study introduces NMRformer, a Transformer-based deep learning framework for accurate peak assignment and metabolite identification in 1D 1H NMR spectroscopy. Unlike traditional approaches, NMRformer interprets spectra as sequences of spectral peaks and integrates a self-attention mechanism and peak height ratios directly into the Transformer encoder layer. It has the capability to recognize and interpret long-range dependencies between peaks and to quickly identify peaks corresponding to identical metabolites. The effectiveness of NMRformer has been rigorously validated by analyzing real 1D 1H NMR spectra from a variety of cellular and biofluid samples. NMRformer achieved peak assignment accuracies above 88% and metabolite identification accuracies above 80% in four types of cellular samples. It also achieved peak assignment accuracies above 88% and metabolite identification accuracies above 80% in three types of biofluid samples. These results underscore the ability of NMRformer to significantly improve the accuracy and efficiency of peak assignment and metabolite identification in NMR-based metabolomics studies.

PMID:39745381 | DOI:10.1021/acs.analchem.4c05632

Soft Matter. 2025 Jan 2. doi: 10.1039/d4sm01160k. Online ahead of print.

ABSTRACT

Monolayer assembly of charged colloidal particles at liquid interfaces opens a new avenue for advancing the additive manufacturing of thin film materials and devices with tailored properties. In this study, we investigated the dynamics of electrosprayed colloidal particles at curved droplet interfaces through a combination of physics-based computational simulations and machine learning. We employed a novel mesh-constrained Brownian dynamics (BD) algorithm coupled with Ansys® electric field simulations to model the transport and assembly of charged particles on a non-spherical droplet surface. We demonstrated that the electrostatic repulsion between particles, electrophoretic forces induced by substrate surface charge, and Brownian motion are the key factors influencing the compactness and ordering of the assembly structure. We further trained a deep neural network surrogate model using the data generated from the BD simulations to predict radial distribution functions (RDF) of particle assembly. By coupling the surrogate model with Bayesian optimization, we identified the optimal particle and substrate charge densities that yield the best match between the simulation and experimental assembly. Using the optimal charge densities, the RDF profile of the simulated assembly accurately matches the experiment with a similarity of 96.4%, and the corresponding average bond order parameter differs by less than 5% from the experimental one. This deep-learning-based approach significantly reduces computational time while maintaining high accuracy in predicting the important features of the assembly structures. The charge densities inferred from the modeling provide critical insights into the surface charge accumulation in the electrospray process.

PMID:39745220 | DOI:10.1039/d4sm01160k

Orthod Craniofac Res. 2025 Jan 2. doi: 10.1111/ocr.12890. Online ahead of print.

ABSTRACT

OBJECTIVE: To assess the accuracy of three commercially available and one open-source deep learning (DL) solutions for automatic tooth segmentation in cone beam computed tomography (CBCT) images of patients with multiple dental impactions.

MATERIALS AND METHODS: Twenty patients (20 CBCT scans) were selected from a retrospective cohort of individuals with multiple dental impactions. For each CBCT scan, one reference segmentation and four DL segmentations of the maxillary and mandibular teeth were obtained. Reference segmentations were generated by experts using a semi-automatic process. DL segmentations were automatically generated according to the manufacturer's instructions. Quantitative and qualitative evaluations of each DL segmentation were performed by comparing it with expert-generated segmentation. The quantitative metrics used were Dice similarity coefficient (DSC) and the normalized surface distance (NSD).

RESULTS: The patients had an average of 12 retained teeth, with 12 of them diagnosed with a rare disease. DSC values ranged from 88.5% ± 3.2% to 95.6% ± 1.2%, and NSD values ranged from 95.3% ± 2.7% to 97.4% ± 6.5%. The number of completely unsegmented teeth ranged from 1 (0.1%) to 41 (6.0%). Two solutions (Diagnocat and DentalSegmentator) outperformed the others across all tested parameters.

CONCLUSION: All the tested methods showed a mean NSD of approximately 95%, proving their overall efficiency for tooth segmentation. The accuracy of the methods varied among the four tested solutions owing to the presence of impacted teeth in our CBCT scans. DL solutions are evolving rapidly, and their future performance cannot be predicted based on our results.

PMID:39744906 | DOI:10.1111/ocr.12890

J Drug Target. 2025 Jan 2:1-85. doi: 10.1080/1061186X.2024.2448711. Online ahead of print.

ABSTRACT

A significant area of computer science called artificial intelligence (AI) is successfully applied to the analysis of intricate biological data and the extraction of substantial associations from datasets for a variety of biomedical uses. AI has attracted significant interest in biomedical research due to its features: (i) better patient care through early diagnosis and detection; (ii) enhanced workflow; (iii) lowering medical errors; (v) lowering medical costs; (vi) reducing morbidity and mortality; (vii) enhancing performance; (viii) enhancing precision; and (ix) time efficiency. Quantitative metrics are crucial for evaluating AI implementations, providing insights, enabling informed decisions, and measuring the impact of AI-driven initiatives, thereby enhancing transparency, accountability, and overall impact. The implementation of AI in biomedical fields faces challenges such as ethical and privacy concerns, lack of awareness, technology unreliability, and professional liability. A brief discussion is given of the AI techniques, which include Virtual screening (VS), DL, ML, Hidden Markov models (HMMs), Neural networks (NNs), Generative models (GMs), Molecular dynamics (MD), and Structure-activity relationship (SAR) models. The study explores the application of AI in biomedical fields, highlighting its enhanced predictive accuracy, treatment efficacy, diagnostic efficiency, faster decision-making, personalized treatment strategies, and precise medical interventions.

PMID:39744873 | DOI:10.1080/1061186X.2024.2448711

Laryngoscope. 2025 Jan 2. doi: 10.1002/lary.31979. Online ahead of print.

ABSTRACT

OBJECTIVE: Automatic segmentation and detection of vestibular schwannoma (VS) in MRI by deep learning is an upcoming topic. However, deep learning faces generalization challenges due to tumor variability even though measurements and segmentation of VS are essential for growth monitoring and treatment planning. Therefore, we introduce a novel model combining two Convolutional Neural Network (CNN) models for the detection of VS by deep learning aiming to improve performance of automatic segmentation.

METHODS: Deep learning techniques have been employed for automatic VS tumor segmentation, including 2D, 2.5D, and 3D UNet-like architectures, which is a specific CNN designed to improve automatic segmentation for medical imaging. Specifically, we introduce a sequential connection where the first UNet's predicted segmentation map is passed to a second complementary network for refinement. Additionally, spatial attention mechanisms are utilized to further guide refinement in the second network.

RESULTS: We conducted experiments on both public and private datasets containing contrast-enhanced T1 and high-resolution T2-weighted magnetic resonance imaging (MRI). Across the public dataset, we observed consistent improvements in Dice scores for all variants of 2D, 2.5D, and 3D CNN methods, with a notable enhancement of 8.86% for the 2D UNet variant on T1. In our private dataset, a 3.75% improvement was reported for 2D T1. Moreover, we found that T1 images generally outperformed T2 in VS segmentation.

CONCLUSION: We demonstrate that sequential connection of UNets combined with spatial attention mechanisms enhances VS segmentation performance across state-of-the-art 2D, 2.5D, and 3D deep learning methods.

LEVEL OF EVIDENCE: 3 Laryngoscope, 2024.

PMID:39744768 | DOI:10.1002/lary.31979

World J Orthop. 2024 Dec 18;15(12):1146-1154. doi: 10.5312/wjo.v15.i12.1146. eCollection 2024 Dec 18.

ABSTRACT

BACKGROUND: Pes planus (flatfoot) and pes cavus (high arch foot) are common foot deformities, often requiring clinical and radiographic assessment for diagnosis and potential subsequent management. Traditional diagnostic methods, while effective, pose limitations such as cost, radiation exposure, and accessibility, particularly in underserved areas.

AIM: To develop deep learning algorithms that detect and classify such deformities using smartphone cameras.

METHODS: An algorithm that integrated a deep convolutional neural network (CNN) into a smartphone camera was utilized to detect pes planus and pes cavus deformities. This case control study was conducted at a tertiary hospital with participants recruited from two orthopaedic foot and ankle clinics. The CNN was trained and tested using photographs of the medial aspect of participants' feet, taken under standardized conditions. Participants included subjects with standard foot alignment, pes planus, or pes cavus determined by an expert clinician using the foot posture index. The model's performance was assessed in comparison to clinical assessment and radiographic measurements, specifically lateral tarsal-first metatarsal angle and calcaneal inclination angle.

RESULTS: The CNN model demonstrated high accuracy in diagnosing both pes planus and pes cavus, with an optimized area under the curve of 0.90 for pes planus and 0.90 for pes cavus. It showed a specificity and sensitivity of 84% and 87% for pes planus detection, respectively; and 97% and 70% for pes cavus, respectively. The model's prediction correlated moderately with radiographic lateral Meary's angle measurements, indicating the model's excellent reliability in assessing food arch deformity (P < 0.05).

CONCLUSION: This study highlights the potential of using a smartphone-based CNN model as a screening tool that is reliable and accessible for the detection of pes planus and pes cavus deformities, which is especially beneficial for underserved communities and patients with pain generated by subtle foot arch deformities.

PMID:39744730 | PMC:PMC11686530 | DOI:10.5312/wjo.v15.i12.1146

Front Comput Neurosci. 2024 Dec 18;18:1468519. doi: 10.3389/fncom.2024.1468519. eCollection 2024.

ABSTRACT

It is a universal phenomenon for patients who do not know which clinical department to register in large general hospitals. Although triage nurses can help patients, due to the larger number of patients, they have to stand in a queue for minutes to consult. Recently, there have already been some efforts to devote deep-learning techniques or pre-trained language models (PLMs) to triage recommendations. However, these methods may suffer two main limitations: (1) These methods typically require a certain amount of labeled or unlabeled data for model training, which are not always accessible and costly to acquire. (2) These methods have not taken into account the distortion of semantic feature structure and the loss of category discriminability in the model training. To overcome these limitations, in this study, we propose a cross-domain text classification method based on prompt-tuning, which can classify patients' questions or texts about their symptoms into several given categories to give suggestions on which kind of consulting room patients could choose. Specifically, first, different prompt templates are manually crafted based on various data contents, embedding source domain information into the prompt templates to generate another text with similar semantic feature structures for performing classification tasks. Then, five different strategies are employed to expand the label word space for modifying prompts, and the integration of these strategies is used as the final verbalizer. The extensive experiments on Chinese Triage datasets demonstrate that our method achieved state-of-the-art performance.

PMID:39744724 | PMC:PMC11688176 | DOI:10.3389/fncom.2024.1468519

Front Comput Neurosci. 2024 Dec 18;18:1505746. doi: 10.3389/fncom.2024.1505746. eCollection 2024.

ABSTRACT

BACKGROUND: Automatic sleep staging is essential for assessing sleep quality and diagnosing sleep disorders. While previous research has achieved high classification performance, most current sleep staging networks have only been validated in healthy populations, ignoring the impact of Obstructive Sleep Apnea (OSA) on sleep stage classification. In addition, it remains challenging to effectively improve the fine-grained detection of polysomnography (PSG) and capture multi-scale transitions between sleep stages. Therefore, a more widely applicable network is needed for sleep staging.

METHODS: This paper introduces MSDC-SSNet, a novel deep learning network for automatic sleep stage classification. MSDC-SSNet transforms two channels of electroencephalogram (EEG) and one channel of electrooculogram (EOG) signals into time-frequency representations to obtain feature sequences at different temporal and frequency scales. An improved Transformer encoder architecture ensures temporal consistency and effectively captures long-term dependencies in EEG and EOG signals. The Multi-Scale Feature Extraction Module (MFEM) employs convolutional layers with varying dilation rates to capture spatial patterns from fine to coarse granularity. It adaptively fuses the weights of features to enhance the robustness of the model. Finally, multiple channel data are integrated to address the heterogeneity between different modalities effectively and alleviate the impact of OSA on sleep stages.

RESULTS: We evaluated MSDC-SSNet on three public datasets and our collection of PSG records of 17 OSA patients. It achieved an accuracy of 80.4% on the OSA dataset. It also outperformed the state-of-the-art methods in terms of accuracy, F1 score, and Cohen's Kappa coefficient on the remaining three datasets.

CONCLUSION: The MSDC-SSRNet multi-channel sleep staging architecture proposed in this study enhances widespread system applicability by supplementing inter-channel features. It employs multi-scale attention to extract transition rules between sleep stages and effectively integrates multimodal information. Our method address the limitations of single-channel approaches, enhancing interpretability for clinical applications.

PMID:39744723 | PMC:PMC11688327 | DOI:10.3389/fncom.2024.1505746

Front Plant Sci. 2024 Dec 18;15:1434222. doi: 10.3389/fpls.2024.1434222. eCollection 2024.

ABSTRACT

Few-shot learning (FSL) methods have made remarkable progress in the field of plant disease recognition, especially in scenarios with limited available samples. However, current FSL approaches are usually limited to a restrictive setting where base classes and novel classes come from the same domain such as PlantVillage. Consequently, when the model is generalized to new domains (field disease datasets), its performance drops sharply. In this work, we revisit the cross-domain performance of existing FSL methods from both data and model perspectives, aiming to better achieve cross-domain generalization of disease by exploring inter-domain correlations. Specifically, we propose a broader cross-domain few-shot learning(CD-FSL) framework for crop disease identification that allows the classifier to generalize to previously unseen categories and domains. Within this framework, three representative CD-FSL models were developed by integrating the Brownian distance covariance (BCD) module and improving the general feature extractor, namely metric-based CD-FSL(CDFSL-BDC), optimization-based CD-FSL(CDFSL-MAML), and non-meta-learning-based CD-FSL (CDFSL-NML). To capture the impact of domain shift on model performance, six public datasets with inconsistent feature distributions between domains were selected as source domains. We provide a unified testbed to conduct extensive meta-training and meta-testing experiments on the proposed benchmarks to evaluate the generalization performance of CD-FSL in the disease domain. The results showed that the accuracy of the three CD-FSL models improved significantly as the inter-domain similarity increased. Compared with other state-of-the-art CD-FSL models, the CDFSL-BDC models had the best average performance under different domain gaps. Shifting from the pest domain to the crop disease domain, the CDFSL-BDC model achieved an accuracy of 63.95% and 80.13% in the 1-shot/5-shot setting, respectively. Furthermore, extensive evaluation on a multi-domain datasets demonstrated that multi-domain learning exhibits stronger domain transferability compared to single-domain learning when there is a large domain gap between the source and target domain. These comparative results suggest that optimizing the CD-FSL method from a data perspective is highly effective for solving disease identification tasks in field environments. This study holds promise for expanding the application of deep learning techniques in disease detection and provides a technical reference for cross-domain disease detection.

PMID:39744608 | PMC:PMC11688362 | DOI:10.3389/fpls.2024.1434222

Imaging Sci Dent. 2024 Dec;54(4):305-312. doi: 10.5624/isd.20240321. Epub 2024 Aug 25.

ABSTRACT

PURPOSE: This review aimed to explore the scientific literature concerning the methodologies and applications of artificial intelligence (AI) in the field of endodontics. The findings may equip dentists with the necessary technical knowledge to understand the opportunities presented by AI.

MATERIALS AND METHODS: Articles published between 1992 and 2023 were retrieved through an electronic search of Medline via the PubMed, Scopus, and Google Scholar databases. The search, which was limited to articles published in English, aimed to identify relevant studies by employing the following keywords: "artificial intelligence," "machine learning," "deep learning," "endodontic," "root canal treatment," and "radiography." Ultimately, 71 studies that addressed the application of AI in endodontics were selected.

RESULTS: Numerous studies have demonstrated the effectiveness of AI applications in endodontics. These uses encompass the identification of root fractures and periapical lesions, assessment of working length, investigation of root canal system anatomy, prediction of retreatment success, and evaluation of dental pulp stem cell viability.

CONCLUSION: AI technology is poised to advance aspects of endodontics including scheduling, patient care, management of drug-drug interactions, prognostic diagnosis, and the emerging area of robotic endodontic surgery. AI methods have demonstrated accuracy and precision in the identification, assessment, and prediction of diseases. Thus, AI can significantly improve endodontic diagnosis and treatment, increasing the overall efficacy of endodontic therapy.

PMID:39744558 | PMC:PMC11685306 | DOI:10.5624/isd.20240321

Front Mol Neurosci. 2024 Dec 18;17:1497735. doi: 10.3389/fnmol.2024.1497735. eCollection 2024.

ABSTRACT

INTRODUCTION: The prostate is densely innervated like many visceral organs and glands. However, studies to date have focused on sympathetic and parasympathetic nerves and little attention has been given to the presence or function of sensory nerves in the prostate. Recent studies have highlighted a role for sensory nerves beyond perception of noxious stimuli, as anterograde release of neuropeptides from sensory nerves can affect vascular tone and local immune responses.

METHODS: To identify the degree of sensory innervation in the prostate, we utilized state-of-the-art tissue clearing and microscopy to visualize sensory innervation in the different lobes of the mouse prostate. To determine whether sensory nerves have a role in regulating proliferation within the prostate, we used an intersectional genetic and toxin approach to ablate peptidergic sensory nerves systemically.

RESULTS: We found that sensory neurons are abundant in the prostate both in nerve bundles along the vasculature and as independent nerve fibers wrapped around prostatic acini in a net-like fashion. In addition to the dense innervation of the prostate, we found that Calca haploinsufficiency, the genotype control for our intersectional ablation model, results in a diminished level of Ki67 staining in the stromal compartment of the dorsal lobe and a diminishing Ki67 trend in other lobes.

DISCUSSION: These findings suggest that sensory neurons might have developmental or homeostatic effects within the prostate. Further studies are warranted to assess the role of sensory neurons and the sensory neuropeptides on prostatic development and on proliferation in the presence of pro-inflammatory stimuli such as bacterial infection or tumor cells.

PMID:39744541 | PMC:PMC11688385 | DOI:10.3389/fnmol.2024.1497735

Front Med (Lausanne). 2024 Dec 18;11:1521712. doi: 10.3389/fmed.2024.1521712. eCollection 2024.

ABSTRACT

OBJECTIVE: With the development of ChatGPT, the number of studies within the nursing field has increased. The sophisticated language capabilities of ChatGPT, coupled with its exceptional precision, offer significant support within the nursing field, which includes clinical nursing, nursing education, and the clinical decision-making process. Preliminary findings suggest positive outcomes, underscoring its potential as a valuable resource for enhancing clinical care. However, a comprehensive analysis of this domain is lacking, and the application of bibliometric methods remains rare. This study aims to describe and predict the developmental trajectory of the discipline, identify research hotspots and trends, and provide a comprehensive framework for the integration of ChatGPT in nursing.

METHODS: Following the development of a search strategy in collaboration with librarians, the implementation of this strategy occurred in the Web of Science Core Collection (WoSCC) on June 30, 2024. For bibliometric and visual analyses-including evaluations of sources, institutions, countries, author collaboration networks, and keywords-Bibliometrix (version 4.4.2) and CiteSpace (version 6.2.R2 Basic) were employed.

RESULTS: A total of 81 articles published by 67 authors were retrieved from the Web of Science Core Collection database, covering the period of June 30, 2024. The number of published studies has exhibited an increasing trend. The "European Journal of Cardiovascular Nursing" emerged as the most productive journals, while the USA, the UK, and China were identified as the leading countries in terms of publication output. The top 10 keywords identified in this study include artificial intelligence, nursing education, large language models, ChatGPT, natural language processing, generative artificial intelligence, care, nursing practice, clinical decision-making, and deep learning.

CONCLUSION: ChatGPT is an emerging tool in the nursing field, currently in the foundational research phase. While there is significant international collaboration, cooperation among author groups remains somewhat limited. Studies focusing on ChatGPT in nursing primarily concentrate on two key themes: (1) the deep learning of ChatGPT in nursing and (2) the feasibility of its application. It is essential for nurses across various specialties to collaborate in exploring the diverse applications of ChatGPT within their domains, thereby fostering the ongoing development and enhancement of this technology.

PMID:39744533 | PMC:PMC11688491 | DOI:10.3389/fmed.2024.1521712

Eur J Cancer. 2024 Dec 26;216:115199. doi: 10.1016/j.ejca.2024.115199. Online ahead of print.

ABSTRACT

PURPOSE: Ovarian cancer patients with a Homologous Recombination Deficiency (HRD) often benefit from polyadenosine diphosphate-ribose polymerase (PARP) inhibitor maintenance therapy after response to platinum-based chemotherapy. HR status is currently analyzed via complex molecular tests. Predicting benefit from PARP inhibitors directly on histological whole slide images (WSIs) could be a fast and cheap alternative.

PATIENTS AND METHODS: We trained a Deep Learning (DL) model on H&E stained WSIs with "shrunken centroid" (SC) based HRD ground truth using the AGO-TR1 cohort (n = 208: 108 training, 100 test) and tested its ability to predict HRD as evaluated by the Myriad classifier and the benefit from olaparib in the PAOLA-1 cohort (n = 447) in a blinded manner.

RESULTS: In contrast to the HRD prediction AUROC of 72 % on hold-out, our model only yielded an AUROC of 57 % external. Kaplan-Meier analysis showed that progression free survival (PFS) in the PARP inhibitor treated PAOLA-1 patients was significantly improved in the HRD positive group as defined by our model, but not in the HRD negative group. PFS improvement in PARP inhibitor-treated patients was substantially longer in our HRD positive group, hinting at a biologically meaningful prediction of benefit from PARP inhibitors.

CONCLUSION: Together, our results indicate that it might be possible to generate a predictor of benefit from PARP inhibitors based on the DL-mediated analysis of WSIs. However, further studies with larger cohorts and further methodological improvements will be necessary to generate a predictor with clinically useful accuracy across independent patient cohorts.

PMID:39742559 | DOI:10.1016/j.ejca.2024.115199

Sci Adv. 2025 Jan 3;11(1):eadn1870. doi: 10.1126/sciadv.adn1870. Epub 2025 Jan 1.

ABSTRACT

DNA methylation (DNAm) is essential for brain development and function and potentially mediates the effects of genetic risk variants underlying brain disorders. We present INTERACT, a transformer-based deep learning model to predict regulatory variants affecting DNAm levels in specific brain cell types, leveraging existing single-nucleus DNAm data from the human brain. We show that INTERACT accurately predicts cell type-specific DNAm profiles, achieving an average area under the receiver operating characteristic curve of 0.99 across cell types. Furthermore, INTERACT predicts cell type-specific DNAm regulatory variants, which reflect cellular context and enrich the heritability of brain-related traits in relevant cell types. We demonstrate that incorporating predicted variant effects and DNAm levels of CpG sites enhances the fine mapping for three brain disorders-schizophrenia, depression, and Alzheimer's disease-and facilitates mapping causal genes to particular cell types. Our study highlights the power of deep learning in identifying cell type-specific regulatory variants, which will enhance our understanding of the genetics of complex traits.

PMID:39742481 | DOI:10.1126/sciadv.adn1870

NPJ Syst Biol Appl. 2025 Jan 2;11(1):1. doi: 10.1038/s41540-024-00484-9.

ABSTRACT

CITE-seq provides a powerful method for simultaneously measuring RNA and protein expression at the single-cell level. The integrated analysis of RNA and protein expression in identical cells is crucial for revealing cellular heterogeneity. However, the high experimental costs associated with CITE-seq limit its widespread application. In this paper, we propose scTEL, a deep learning framework based on Transformer encoder layers, to establish a mapping from sequenced RNA expression to unobserved protein expression in the same cells. This computation-based approach significantly reduces the experimental costs of protein expression sequencing. We are now able to predict protein expression using single-cell RNA sequencing (scRNA-seq) data, which is well-established and available at a lower cost. Moreover, our scTEL model offers a unified framework for integrating multiple CITE-seq datasets, addressing the challenge posed by the partial overlap of protein panels across different datasets. Empirical validation on public CITE-seq datasets demonstrates scTEL significantly outperforms existing methods.

PMID:39743530 | DOI:10.1038/s41540-024-00484-9

Methods. 2024 Dec 30:S1046-2023(24)00284-6. doi: 10.1016/j.ymeth.2024.12.010. Online ahead of print.

ABSTRACT

5-Methylcytosine (m5C) plays a pivotal role in various RNA metabolic processes, including RNA localization, stability, and translation. Current high-throughput sequencing technologies for m5C site identification are resource-intensive in terms of cost, labor, and time. As such, there is a pressing need for efficient computational approaches. Many existing computational methods rely on intricate hand-crafted features, requiring unavailable features, often leading to suboptimal prediction accuracy. Addressing these challenges, we introduce a novel deep-learning method, Trans-m5C. We first categorize m5C sites into NSUN2-dependent and NSUN6-dependent types for independent feature extraction. Subsequently, meticulously crafted transformer neural networks are employed to distill global features. The prediction of m5C sites is then accomplished using a discriminator built from a a multi-layer perceptron. A rigorous evaluation for the performance of Trans-m5C on experimentally validated m5C data from human and mouse species reveals that our method offers a competitive edge over both baseline and existing methodologies.

PMID:39742984 | DOI:10.1016/j.ymeth.2024.12.010

Environ Int. 2024 Dec 28;195:109244. doi: 10.1016/j.envint.2024.109244. Online ahead of print.

ABSTRACT

Chemically induced neurotoxicity is a critical aspect of chemical safety assessment. Traditional and costly experimental methods call for the development of high-throughput virtual screening. However, the small datasets of neurotoxicity have limited the application of advanced deep learning techniques. The current study developed a hybrid deep learning architecture, NeuTox 2.0, through multimodal feature fusion for enhanced prediction accuracy and generalization ability. We incorporated transfer learning based on self-supervised learning, graph neural networks, and molecular fingerprints/descriptors. Four datasets were used to profile neurotoxicity; these were related to blood-brain barrier permeability, neuronal cytotoxicity, microelectrode array-based neural activity, and mammalian neurotoxicity. Comprehensive performance evaluations demonstrated that NeuTox 2.0 has relatively higher predictive capability across all statistical metrics. Specifically, NeuTox 2.0 exhibits remarkable performance in three of the four datasets. In the BBB dataset, although it does not outperform the PaDEL descriptor model, its performance closely approximates that of the top single-modal model. The ablation experiments indicated that NeuTox 2.0 can learn the deeper structural differences of molecules from various feature extractions and capture complex interactions and mapping relationships between various modalities, thereby improving performance for neurotoxicity prediction. Evaluations of anti-noise ability indicated that NeuTox 2.0 has excellent noise resistance relative to traditional machine learning. We applied the NeuTox 2.0 model to predict the neurotoxicity of 315,790 compounds in the REACH database. The results showed that 701 compounds exhibited potential neurotoxicity in the four neurotoxicity-related predictions. In conclusion, NeuTox 2.0 can be used as an efficient tool for early neurotoxicity screening of environmental chemicals.

PMID:39742830 | DOI:10.1016/j.envint.2024.109244

Comput Biol Med. 2024 Dec 31;186:109641. doi: 10.1016/j.compbiomed.2024.109641. Online ahead of print.

ABSTRACT

Osteoarthritis (OA) is a prevalent condition resulting in physical limitations. Early detection of OA is critical to effectively manage this condition. However, the diagnosis of early-stage arthritis remains challenging. The Kellgren and Lawrence (KL) grading system is a common method that is accepted worldwide, uses five grades to classify the severity of OA, and relies on the ability of the orthopedist to accurately interpret radiograph images. To improve the accuracy of radiograph image interpretation, artificial intelligence-assisted models have been developed that include shallow or deep learning approaches and multi-step techniques; however, their accuracy remains variable. This work proposes a transfer learning approach using an InceptionV3-based model fine-tuned on the Osteoarthritis Initiative dataset, and aims to enhance the identification of OA severity levels through dual-stage preprocessing and convolutional neural networks for feature extraction. The fine-tuned IV3 (FT-IV3) model outperformed the IV3 model with training, validation, and testing accuracies of (96.33, 93.82, and 92.25) %, compared to IV3 accuracies of (91.64, 82.04, and 86.20) %, respectively. Additionally, Cohen's Kappa value for the FT-IV3 model (90.69 %) exceeds that of the IV3 model (83.15 %), indicating a better diagnosis of OA severity. This improvement allows the FT-IV3 model to effectively classify moderate and severe-grade OA.

PMID:39742824 | DOI:10.1016/j.compbiomed.2024.109641