Nature. 2025 Feb 26. doi: 10.1038/s41586-025-08636-5. Online ahead of print.

ABSTRACT

The regulation of metabolism is vital to any organism and can be achieved by transcriptionally activating or repressing metabolic genes1-3. Although many examples of transcriptional metabolic rewiring have been reported4, a systems-level study of how metabolism is rewired in response to metabolic perturbations is lacking in any animal. Here we apply Worm Perturb-Seq (WPS)-a high-throughput method combining whole-animal RNA-interference and RNA-sequencing5-to around 900 metabolic genes in the nematode Caenorhabditis elegans. We derive a metabolic gene regulatory network (mGRN) in which 385 perturbations are connected to 9,414 genes by more than 110,000 interactions. The mGRN has a highly modular structure in which 22 perturbation clusters connect to 44 gene expression programs. The mGRN reveals different modes of transcriptional rewiring from simple reaction and pathway compensation to rerouting and more complex network coordination. Using metabolic network modelling, we identify a design principle of transcriptional rewiring that we name the compensation-repression (CR) model. The CR model explains most transcriptional responses in metabolic genes and reveals a high level of compensation and repression in five core metabolic functions related to energy and biomass. We provide preliminary evidence that the CR model may also explain transcriptional metabolic rewiring in human cells.

PMID:40011787 | DOI:10.1038/s41586-025-08636-5

Nature. 2025 Feb 26. doi: 10.1038/s41586-025-08656-1. Online ahead of print.

ABSTRACT

Substantial epigenetic resetting during early embryo development from fertilization to blastocyst formation ensures zygotic genome activation and leads to progressive cellular heterogeneities1-3. Mapping single-cell epigenomic profiles of core histone modifications that cover each individual cell is a fundamental goal in developmental biology. Here we develop target chromatin indexing and tagmentation (TACIT), a method that enabled genome-coverage single-cell profiling of seven histone modifications across mouse early embryos. We integrated these single-cell histone modifications with single-cell RNA sequencing data to chart a single-cell resolution epigenetic landscape. Multimodal chromatin-state annotations showed that the onset of zygotic genome activation at the early two-cell stage already primes heterogeneities in totipotency. We used machine learning to identify totipotency gene regulatory networks, including stage-specific transposable elements and putative transcription factors. CRISPR activation of a combination of these identified transcription factors induced totipotency activation in mouse embryonic stem cells. Together with single-cell co-profiles of multiple histone modifications, we developed a model that predicts the earliest cell branching towards the inner cell mass and the trophectoderm in latent multimodal space and identifies regulatory elements and previously unknown lineage-specifying transcription factors. Our work provides insights into single-cell epigenetic reprogramming, multimodal regulation of cellular lineages and cell-fate priming during mouse pre-implantation development.

PMID:40011786 | DOI:10.1038/s41586-025-08656-1

Nature. 2025 Feb 26. doi: 10.1038/s41586-025-08635-6. Online ahead of print.

ABSTRACT

Metabolic flux, or the rate of metabolic reactions, is one of the most fundamental metrics describing the status of metabolism in living organisms. However, measuring fluxes across the entire metabolic network remains nearly impossible, especially in multicellular organisms. Computational methods based on flux balance analysis have been used with genome-scale metabolic network models to predict network-level flux wiring1-6. However, such approaches have limited power because of the lack of experimental constraints. Here, we introduce a strategy that infers whole-animal metabolic flux wiring from transcriptional phenotypes in the nematode Caenorhabditis elegans. Using a large-scale Worm Perturb-Seq (WPS) dataset for roughly 900 metabolic genes7, we show that the transcriptional response to metabolic gene perturbations can be integrated with the metabolic network model to infer a highly constrained, semi-quantitative flux distribution. We discover several features of adult C. elegans metabolism, including cyclic flux through the pentose phosphate pathway, lack of de novo purine synthesis flux and the primary use of amino acids and bacterial RNA as a tricarboxylic acid cycle carbon source, all of which we validate by stable isotope tracing. Our strategy for inferring metabolic wiring based on transcriptional phenotypes should be applicable to a variety of systems, including human cells.

PMID:40011784 | DOI:10.1038/s41586-025-08635-6

Nat Rev Gastroenterol Hepatol. 2025 Feb 26. doi: 10.1038/s41575-025-01042-2. Online ahead of print.

ABSTRACT

The demonstration that Helicobacter pylori is a pathogenic bacterium with marked carcinogenic potential has paved the way for new preventive approaches for gastric cancer. Although decades of research have uncovered complex interactions of H. pylori with epithelial cells, current insights have refined our view on H. pylori-associated carcinogenesis. Specifically, the cell-type-specific effects on gastric stem and progenitor cells deep in gastric glands provide a new view on the ability of the bacteria to colonize long-term, manipulate host responses and promote gastric pathology. Furthermore, new, large-scale epidemiological data have shed light on factors that determine why only a subset of carriers progress to gastric cancer. Currently, technological advances have brought yet another revelation: H. pylori is far from the only microorganism able to colonize the stomach. Instead, the stomach is colonized by a diverse gastric microbiota, and there is emerging evidence for the occurrence and pathological effect of dysbiosis resulting from an aberrant interplay between H. pylori and the gastric mucosa. With the weight of this evidence mounting, here we consider how the lessons learned from H. pylori research inform and synergize with this emerging field to bring a more comprehensive understanding of the role of microbes in gastric carcinogenesis.

PMID:40011753 | DOI:10.1038/s41575-025-01042-2

Nat Microbiol. 2025 Feb 26. doi: 10.1038/s41564-025-01935-7. Online ahead of print.

ABSTRACT

Bacteriophages constitute one of the largest reservoirs of genes of unknown function in the biosphere. Even in well-characterized phages, the functions of most genes remain unknown. Experimental approaches to study phage gene fitness and function at genome scale are lacking, partly because phages subvert many modern functional genomics tools. Here we leverage RNA-targeting dCas13d to selectively interfere with protein translation and to measure phage gene fitness at a transcriptome-wide scale. We find CRISPR Interference through Antisense RNA-Targeting (CRISPRi-ART) to be effective across phage phylogeny, from model ssRNA, ssDNA and dsDNA phages to nucleus-forming jumbo phages. Using CRISPRi-ART, we determine a conserved role of diverse rII homologues in subverting phage Lambda RexAB-mediated immunity to superinfection and identify genes critical for phage fitness. CRISPRi-ART establishes a broad-spectrum phage functional genomics platform, revealing more than 90 previously unknown genes important for phage fitness.

PMID:40011704 | DOI:10.1038/s41564-025-01935-7

Nat Med. 2025 Feb 26. doi: 10.1038/s41591-025-03533-w. Online ahead of print.

ABSTRACT

The role of endothelial cell (EC) dysfunction in contributing to an individual's susceptibility to coronary atherosclerosis and how low-density lipoprotein cholesterol (LDL-C) concentrations might modify this relationship have not been previously studied. Here, from an examination of genome-wide significant single nucleotide polymorphisms associated with coronary artery disease (CAD), we identified variants with effects on EC function and constructed a 35 single nucleotide polymorphism polygenic risk score comprising these EC-specific variants (EC PRS). The association of the EC PRS with the risk of incident cardiovascular disease was tested in 3 cohorts: a primary prevention population in the UK Biobank (UKBB; n = 348,967); a primary prevention cohort from a trial that tested a statin (JUPITER, n = 8,749); and a secondary prevention cohort that tested a PCSK9 inhibitor (FOURIER, n = 14,298). In the UKBB, the EC PRS was independently associated with the risk of incident CAD (adjusted hazard ratio (aHR) per 1 s.d. of 1.24 (95% CI 1.21-1.26), P < 2 × 10-16). Moreover, LDL-C concentration significantly modified this risk: the aHR per 1 s.d. was 1.26 (1.22-1.30) when LDL-C was 150 mg dl-1 but 1.00 (0.85-1.16) when LDL-C was 50 mg dl-1 (Pinteraction = 0.004). The clinical benefit of LDL-C lowering was significantly greater in individuals with a high EC PRS than in individuals with low or intermediate EC PRS, with relative risk reductions of 68% (HR 0.32 (0.18-0.59)) versus 29% (HR 0.71 (0.52-0.95)) in the primary prevention cohort (Pinteraction = 0.02) and 33% (HR 0.67 (0.53-0.83)) versus 8% (HR 0.92 (0.82-1.03)) in the secondary prevention cohort (Pinteraction = 0.01). We conclude that EC PRS quantifies an independent axis of CAD risk that is not currently captured in medical practice and identifies individuals who are more sensitive to the atherogenic effects of LDL-C and who would potentially derive substantially greater benefit from aggressive LDL-C lowering.

PMID:40011692 | DOI:10.1038/s41591-025-03533-w

Nat Cell Biol. 2025 Feb 26. doi: 10.1038/s41556-025-01622-z. Online ahead of print.

ABSTRACT

Recent advancements in functional genomics have provided an unprecedented ability to measure diverse molecular modalities, but predicting causal regulatory relationships from observational data remains challenging. Here, we leverage pooled genetic screens and single-cell sequencing (Perturb-seq) to systematically identify the targets of signalling regulators in diverse biological contexts. We demonstrate how Perturb-seq is compatible with recent and commercially available advances in combinatorial indexing and next-generation sequencing, and perform more than 1,500 perturbations split across six cell lines and five biological signalling contexts. We introduce an improved computational framework (Mixscale) to address cellular variation in perturbation efficiency, alongside optimized statistical methods to learn differentially expressed gene lists and conserved molecular signatures. Finally, we demonstrate how our Perturb-seq derived gene lists can be used to precisely infer changes in signalling pathway activation for in vivo and in situ samples. Our work enhances our understanding of signalling regulators and their targets, and lays a computational framework towards the data-driven inference of an 'atlas' of perturbation signatures.

PMID:40011560 | DOI:10.1038/s41556-025-01622-z

Nat Commun. 2025 Feb 26;16(1):1995. doi: 10.1038/s41467-025-57182-1.

ABSTRACT

Hepatitis E virus (HEV) causes 3.3 million symptomatic cases and 44,000 deaths per year. Chronic infections can arise in immunocompromised individuals, and pregnant women may suffer from fulminant disease as a consequence of HEV infection. Despite these important implications for public health, no specific antiviral treatment has been approved to date. Here, we report combined functional, biochemical, and X-ray crystallographic studies that characterize the human antibody response in convalescent HEV patients. We identified a class of potent and broadly neutralizing human antibodies (bnAbs), targeting a quaternary epitope located at the tip of the HEV capsid protein pORF2 that contains an N-glycosylation motif and is conserved across members of the Hepeviridae. These glycan-sensitive bnAbs specifically recognize the non-glycosylated pORF2 present in infectious particles but not the secreted glycosylated form acting as antibody decoy. Our most potent bnAb protects human liver-chimeric mice from intraperitoneal HEV challenge and co-housing exposure. These results provide insights into the bnAb response to this important emerging pathogen and support the development of glycan-sensitive antibodies to combat HEV infection.

PMID:40011441 | DOI:10.1038/s41467-025-57182-1

Life Sci. 2025 Feb 24:123507. doi: 10.1016/j.lfs.2025.123507. Online ahead of print.

ABSTRACT

AIMS: The association between aging-related hyperphosphatemia and sarcopenia has been documented, and evidence suggests that inflammaging is involved in the manifestation of sarcopenia. The present study investigates whether hyperphosphatemia triggers inflammation, thereby inducing the appearance of sarcopenia along with the cytokines involved in these processes.

MATERIALS AND METHODS: RAW 264.7 macrophages were incubated with β-glycerophosphate (BGP), as a phosphate donor, at different time intervals, to assess the production of proinflammatory markers. Conditioned medium from macrophages was collected and added to cultured C2C12 myoblasts to analyse whether proinflammatory molecules, released by macrophages, modified myogenic differentiation, cell senescence or myokine IL-15 expression. A neutralising antibody anti-TNF-α and recombinant IL-15 were added to evaluate the role of these cytokines in the observed effects. Additionally, TNF-α, IL-15, serum phosphate, and sarcopenia signs were evaluated in 5-month-old mice, 24-month-old mice and 24-month-old mice fed with a hypophosphatemic diet.

KEY FINDINGS: BGP increased TNF-α expression in macrophages through NFkB activation. Conditioned medium from BGP-treated macrophages impaired myogenic differentiation in differentiating myoblasts and promoted cellular senescence and reduced IL-15 expression in undifferentiated myoblasts. These effects were mediated by TNF-α. Old mice displayed reduced expression of muscle IL-15 and elevated circulating TNF-α, along with increased serum phosphate levels, which correlated with the appearance of sarcopenia indicators. The hypophosphatemic diet prevented these changes in old mice.

SIGNIFICANCE: Hyperphosphatemia induces TNF-α production in macrophages, which contributes to the reduced expression of muscular IL-15. This mechanism may play a role in inducing sarcopenia in elderly mice.

PMID:40010633 | DOI:10.1016/j.lfs.2025.123507

Int J Sports Physiol Perform. 2025 Feb 26:1-4. doi: 10.1123/ijspp.2024-0303. Online ahead of print.

ABSTRACT

BACKGROUND: The role of high-volume low-intensity training for enhancing endurance performance has gained growing interest in recent years. Specifically, so-called "zone 2 training" is currently receiving much attention, and many propose that this is the target intensity at which a large proportion of total endurance training should be performed. However, despite the popularity of this concept, there is no clear consensus among coaches, athletes, and scientists regarding the definition of zone 2 training.

PURPOSE: This commentary summarizes the perspectives, experience, and knowledge of an expert panel of 14 applied sport scientists and professional coaches with the aim of providing insight and a basis for definitional consensus on zone 2 training. Moreover, potential training strategies at this intensity are proposed, and the expected physiological adaptations when exercising at this intensity and related research gaps are also discussed.

RESULTS: Experts reached consensus that zone 2 training should preferably be performed at intensities located immediately below the first lactate or ventilatory threshold through continuous, variable, or interval-type sessions. Furthermore, experts expected a broad range of central and peripheral adaptations from zone 2 training. These expected adaptations might not be unique to zone 2 and could also be induced with sessions performed at slightly higher and lower intensities.

CONCLUSIONS: This commentary provides practical insight and unified criteria regarding the preferred intensity, duration, and session type for the optimization of zone 2 training based on the perspectives of acknowledged sport scientists and professional coaches.

PMID:40010355 | DOI:10.1123/ijspp.2024-0303

Phytomedicine. 2025 Feb 14;139:156518. doi: 10.1016/j.phymed.2025.156518. Online ahead of print.

ABSTRACT

BACKGROUND: Drug discovery from nature has a long, ethnopharmacologically-based background. Today, natural resources are undeniably vital reservoirs of active molecules or drug leads. Advances in (bio)informatics and computational biology emphasized the role of herbal medicines in the drug discovery pipeline.

PURPOSE: This review summarizes bioinformatic approaches applied in recent drug discovery from nature.

STUDY DESIGN: It examines advancements in molecular networking, pathway analysis, network pharmacology within a systems biology framework and AI for assessing the therapeutic potential of herbal preparations.

METHODS: A comprehensive literature search was conducted using Pubmed, SciFinder, and Google Database. Obtained data was analyzed and organized in subsections: AI, systems biology integrative approach, network pharmacology, pathway analysis, molecular networking, structure-based virtual screening.

RESULTS: Bioinformatic approaches is now essential for high-throughput data analysis in drug target identification, mechanism-based drug discovery, drug repurposing and side-effects prediction. Large datasets obtained from "omics" approaches require bioinformatic calculations to unveil interactions, and patterns in disease-relevant conditions. These tools enable databases annotations, pattern-matching, connections discovery, molecular relationship exploration, and data visualisation.

CONCLUSION: Despite the complexity of plant metabolites, bioinformatic approaches assist in characterization of herbal preparations and selection of bioactive molecule. It is perceived as powerful tool for uncovering multi-target effects and potential molecular mechanisms of compounds. By integrating multiple networks that connect gene-disease, drug-target and gene-drug-target, drug discovery from natural sources is experiencing a remarkable comeback.

PMID:40010031 | DOI:10.1016/j.phymed.2025.156518

Eur J Cancer. 2025 Feb 15;219:115304. doi: 10.1016/j.ejca.2025.115304. Online ahead of print.

ABSTRACT

AIM: Concurrent chemoradiotherapy (CCRT) is the standard treatment for locally advanced cervical cancer (LACC), but recurrence rates remain high. This multicenter phase-3 randomized trial (GOTIC-002) evaluated the efficacy of low-dose oral tegafur-uracil (UFT) as maintenance chemotherapy following curative CCRT for LACC.

METHODS: Between 2010 and 2018, 351 patients with stage Ib2-IVa cervical cancer were enrolled. After achieving complete or partial remission post-CCRT, patients were randomized 1:1 into observation (arm O) or UFT maintenance therapy (arm UFT). UFT doses were 300-400 mg/day based on body surface area for 2 years, disease progression or adverse effects occurred. The primary endpoint was progression-free survival (PFS), with overall survival (OS) and safety as secondary endpoints.

RESULTS: Patient characteristics were similar between the groups (n = 178 in arm O, n = 173 in arm UFT). During a median follow-up of 3 years, median PFS was not reached in either group. 5-year PFS rates were similar between them (arm O: 61.3 %, arm UFT: 62.0 %, hazard ratio: 0.92, P = .634). 5-year OS rates were also comparable (77.6 % vs 76.1 %, hazard ratio: 1.04, P = .869). Compliance with UFT ranged from 87.8 % to 98.8 %. Although adverse events were more frequent in arm UFT (93.5 % vs 73.9 %, odds ratio: 5.05), most were mild or moderate. Despite its tolerability, UFT did not improve PFS or OS.

CONCLUSIONS: These findings suggest the need to reconsider maintenance therapy strategies after CCRT for potentially shifting away from cytotoxic chemotherapy towards alternative methods to enhance survival outcomes in patients with LACC.

PMID:40010135 | DOI:10.1016/j.ejca.2025.115304

J Mol Graph Model. 2025 Feb 21;137:108985. doi: 10.1016/j.jmgm.2025.108985. Online ahead of print.

ABSTRACT

Recent advancements in Artificial Intelligence (AI) and deep learning have had a significant impact on drug discovery. The prediction of molecular properties, such as toxicity and blood-brain barrier (BBB) permeability, is crucial for accelerating drug development. The accuracy of these predictions largely depends on the selection of molecular descriptors. Self-supervised learning (SSL) has gained prominence due to its strong generalization capabilities. Graph contrastive learning (GCL), a type of SSL, is particularly useful in this context. Current GCL methods for molecular graphs use various data augmentation techniques, which may potentially alter the inherent structure of molecules. Additionally, traditional single-perspective representations do not fully capture the complexity of molecules. We present RFA-FFM (Fragment-level Feature Fusion Method using Retrosynthetic Fragmentation Algorithm), which integrates molecular representations from multiple perspectives. This method employs two strategies: (1) contrasting chemical information from fragments generated by two retrosynthetic methods to provide detailed contrastive insights; (2) fusing chemical information at different levels of molecular hierarchy, including the entire molecule and its fragments. Experiments show that RFA-FFM enhances the performance of deep learning models in predicting molecular properties, improving ROC-AUC scores by 0.3 %-2.6 % compared to baselines across four classification benchmarks. Case studies on hepatitis B virus datasets demonstrate that RFA-FFM outperforms baselines by 7 %-11 %. When compared to BPE and CC-Single fragmentation algorithms, RFA-FFM shows a 2 %-4 % improvement in BBB permeability tasks, thus demonstrating its effectiveness in predicting molecular properties.

PMID:40009893 | DOI:10.1016/j.jmgm.2025.108985

J Neural Eng. 2025 Feb 26. doi: 10.1088/1741-2552/adba8c. Online ahead of print.

ABSTRACT

OBJECTIVE: Functional brain network (FBN) methods are commonly integrated with deep learning (DL) models for EEG analysis. Typically, an FBN is constructed to extract features from EEG data, which are then fed into a DL model for further analysis. Beyond this two-step approach, there is potential to embed FBN construction directly within DL models as a feature extraction module, enabling the models to learn EEG representations end-to-end while incorporating insights from FBNs. However, a critical prerequisite is whether DL models can effectively learn the FBN construction process.

APPROACH: To address this, we propose using DL models to learn FBN matrices derived from EEG data. The ability of DL models to accurately reproduce these matrices would validate their capacity to learn the FBN construction process. This approach is tested on two publicly available EEG datasets, utilizing seven DL models to learn four representative FBN matrices. Model performance is assessed through mean squared error (MSE), Pearson correlation coefficient (Corr), and concordance correlation coefficient (CCC) between predicted and actual matrices.

MAIN RESULTS: The results show that DL models achieve low MSE and relatively high Corr and CCC values when learning the Coherence network. Visualizations of predicted and error matrices reveal that while DL models capture the general structure of all four FBNs, certain regions remain difficult to model accurately. Additionally, a paired t-test comparing global efficiency and nodal degree between predicted and actual networks indicates that most predicted networks significantly differ from the actual networks (p < 0.05).

SIGNIFICANCE: These findings suggest that while DL&#xD;models can learn the connectivity relationships of certain FBNs, they struggle to capture the intrinsic topological structures. This highlights the irreplaceability of traditional FBN methods in EEG analysis and underscores the need for hybrid strategies that combine FBN methods with DL models for a more comprehensive analysis.

PMID:40009886 | DOI:10.1088/1741-2552/adba8c

J Neural Eng. 2025 Feb 26. doi: 10.1088/1741-2552/adba8a. Online ahead of print.

ABSTRACT

Brain-computer interface (BCI) is a cutting-edge technology that enables interaction with external devices by decoding human intentions, and is highly valuable in the fields of medical rehabilitation and human-robot collaboration. The technique of decoding motor intent for motor execution (ME) based on electroencephalographic (EEG) signals is in the feasibility study stage. There are still insufficient studies on the accuracy of motor execution EEG signal recognition in between-subjects classification to reach the level of realistic applications. This paper aims to investigate EEG signal-driven hand movement recognition by analyzing low-frequency time-domain (LFTD) information. Experiments with four types of hand movements, two force parameter (extraction and pushing) tasks, and a four-target directional displacement task were designed and executed, and the EEG data from thirteen healthy volunteers was collected. Sliding window approach is used to expand the dataset in order to address the issue of EEG signal overfitting. Furtherly, CNN-BiLSTM model, an end-to-end serial combination of a Bidirectional Long Short-Term Memory Network (BiLSTM) and Convolutional Neural Network (CNN) is constructed to classify the raw EEG data to recognize the hand movement. According to experimental data, the model is able to categorize four types of hand movements, extraction movements, pushing movements, and four target direction displacement movements with an accuracy of 99.14%±0.49%, 99.29%±0.11%, 99.23%±0.60%, and 98.11%± 0.23%, respectively. Furthermore, comparative tests conducted with alternative deep learning models (LSTM, CNN, EEGNet, CNN-LSTM) demonstrates that the CNN-BiLSTM model is with practicable accuracy in terms of EEG-based hand movement recognition and its parameter decoding.

PMID:40009879 | DOI:10.1088/1741-2552/adba8a

Invest Radiol. 2025 Feb 25. doi: 10.1097/RLI.0000000000001168. Online ahead of print.

ABSTRACT

OBJECTIVE: The 3-dimensional (3D) double echo steady state (DESS) magnetic resonance imaging sequence can image knee cartilage with high, isotropic resolution, particularly at high and ultra-high field strengths. Advanced undersampling techniques with high acceleration factors can provide the short acquisition times required for clinical use. However, the optimal undersampling scheme and its limits are unknown.

MATERIALS AND METHODS: High-resolution isotropic (reconstructed voxel size: 0.3 × 0.3 × 0.3 mm3) 3D DESS images of 40 knees in 20 volunteers were acquired at 7 T with varying undersampling factors (R = 4-30) and schemes (regular: GRAPPA, CAIPIRINHA; incoherent: compressed sensing [CS]), whereas the remaining imaging parameters were kept constant. All imaging data were reconstructed with deep learning (DL) algorithms. Three readers rated image quality on a 4-point Likert scale. Four-fold accelerated GRAPPA was used as reference standard. Incidental cartilage lesions were graded on a modified Whole-Organ Magnetic Resonance Imaging Score (WORMS). Friedman's analysis of variance characterized rating differences. The interreader agreement was assessed using κ statistics.

RESULTS: The quality of 16-fold accelerated CS images was not rated significantly different from that of 4-fold accelerated GRAPPA and 8-fold accelerated CAIPIRINHA images, whereas the corresponding data were acquired 4.5 and 2 times faster (01:12 min:s) than in 4-fold accelerated GRAPPA (5:22 min:s) and 8-fold accelerated CAIPIRINHA (2:22 min:s) acquisitions, respectively. Interreader agreement for incidental cartilage lesions was almost perfect for 4-fold accelerated GRAPPA (κ = 0.91), 8-fold accelerated CAIPIRINHA (κ = 0.86), and 8- to 16-fold accelerated CS (κ = 0.91).

CONCLUSIONS: Our results suggest significant advantages of incoherent versus regular undersampling patterns for high-resolution 3D DESS cartilage imaging with high acceleration factors. The combination of CS undersampling with DL reconstruction enables fast, isotropic, high-resolution acquisitions without apparent impairment of image quality. Since DESS specific absorption rate values tend to be moderate, CS DESS with DL reconstruction promises potential for high-resolution assessment of cartilage morphology and other musculoskeletal anatomies at 7 T.

PMID:40009727 | DOI:10.1097/RLI.0000000000001168

PLoS One. 2025 Feb 26;20(2):e0318992. doi: 10.1371/journal.pone.0318992. eCollection 2025.

ABSTRACT

In the acquisition of Magnetic Resonance (MR) images shorter scan times lead to higher image noise. Therefore, automatic image denoising using deep learning methods is of high interest. In this work, we concentrate on image denoising of MR images containing line-like structures such as roots or vessels. In particular, we investigate if the special characteristics of these datasets (connectivity, sparsity) benefit from the use of special loss functions for network training. We hereby translate the Perceptual Loss to 3D data by comparing feature maps of untrained networks in the loss function. We tested the performance of untrained Perceptual Loss (uPL) on 3D image denoising of MR images displaying brain vessels (MR angiograms - MRA) and images of plant roots in soil. In this study, 536 MR images of plant roots in soil and 450 MRA images are included. The plant root dataset is split to 380, 80, and 76 images for training, validation, and testing. The MRA dataset is split to 300, 50, and 100 images for training, validation, and testing. We investigate the impact of various uPL characteristics such as weight initialization, network depth, kernel size, and pooling operations on the results. We tested the performance of the uPL loss on four Rician noise levels (1%, 5%, 10%, and 20%) using evaluation metrics such as the Structural Similarity Index Metric (SSIM). Our results are compared with the frequently used L1 loss for different network architectures. We observe, that our uPL outperforms conventional loss functions such as the L1 loss or a loss based on the Structural Similarity Index Metric (SSIM). For MRA images the uPL leads to SSIM values of 0.93 while L1 and SSIM loss led to SSIM values of 0.81 and 0.88, respectively. The uPL network's initialization is not important (e.g. for MR root images SSIM differences of 0.01 occur across initializations, while network depth and pooling operations impact denoising performance slightly more (SSIM of 0.83 for 5 convolutional layers and kernel size 3 vs. 0.86 for 5 convolutional layers and kernel size 5 for the root dataset). We also find that small uPL networks led to better or comparable results than using large networks such as VGG (e.g. SSIM values of 0.93 and 0.90 for a small and a VGG19 uPL network in the MRA dataset). In summary, we demonstrate superior performance of our loss for both datasets, all noise levels, and three network architectures. In conclusion, for images containing line-like structures, uPL is an alternative to other loss functions for 3D image denoising. We observe that small uPL networks have better or equal performance than very large network architectures while requiring lower computational costs and should therefore be preferred.

PMID:40009630 | DOI:10.1371/journal.pone.0318992

PLoS One. 2025 Feb 26;20(2):e0310992. doi: 10.1371/journal.pone.0310992. eCollection 2025.

ABSTRACT

With the dramatic increase in the number of published papers and the continuous progress of deep learning technology, the research on name disambiguation is at a historic peak, the number of paper authors is increasing every year, and the situation of authors with the same name is intensifying, therefore, it is a great challenge to accurately assign the newly published papers to their respective authors. The current mainstream methods for author disambiguation are mainly divided into two methods: feature-based clustering and connection-based clustering, but none of the current mainstream methods can efficiently deal with the author name disambiguation problem, For this reason, this paper proposes the author name ablation method based on the relational graph heterogeneous attention neural network, first extract the semantic and relational information of the paper, use the constructed graph convolutional embedding module to train the splicing to get a better feature representation, and input the constructed network to get the vector representation. As the existing graph heterogeneous neural network can not learn different types of nodes and edge interaction, add multiple attention, design ablation experiments to verify its impact on the network. Finally improve the traditional hierarchical clustering method, combined with the graph relationship and topology, using training vectors instead of distance calculation, can automatically determine the optimal k-value, improve the accuracy and efficiency of clustering. The experimental results show that the average F1 value of this paper's method on the Aminer dataset is 0.834, which is higher than other mainstream methods.

PMID:40009590 | DOI:10.1371/journal.pone.0310992

Biochem Cell Biol. 2025 Feb 26. doi: 10.1139/bcb-2024-0287. Online ahead of print.

ABSTRACT

Estrogen (E2) regulates the differentiation and proliferation of mammary progenitor cells by modulating the transcription of multiple genes. One of the genes that is downregulated by E2 is SOX2, a transcription factor associated with stem and progenitor cells that is overexpressed during breast tumourigenesis. To elucidate the mechanisms underlying E2-mediated SOX2 repression, we investigated epigenome and transcriptome changes following short- and long-term E2 exposure in breast cancer cells. We found that short-term E2 exposure reduces chromatin accessibility at the downstream SOX2 SRR134 enhancer, decreasing SOX2 expression. In contrast, long-term E2 exposure completely represses SOX2 transcription while maintaining accessibility at the SRR124-134 enhancer cluster, keeping it poised for reactivation. This repression was accompanied by widespread epigenome and transcriptome changes associated with commitment towards a more differentiated and less invasive luminal phenotype. Finally, we identified a role for the transcription factor NFIB in this process, suggesting it collaborates with the estrogen receptor to mediate SOX2 repression and genome-wide epigenome accessibility changes.

PMID:40009831 | DOI:10.1139/bcb-2024-0287

Sci Adv. 2025 Feb 28;11(9):eadv2067. doi: 10.1126/sciadv.adv2067. Epub 2025 Feb 26.

ABSTRACT

3D organization of the genome plays a critical role in regulating gene expression. How 3D-genome organization differs among different cell types and relates to cell type-dependent transcriptional regulation remains unclear. Here, we used genome-scale DNA and RNA imaging to investigate 3D-genome organization in transcriptionally distinct cell types in the mouse cerebral cortex. We uncovered a wide spectrum of differences in the nuclear architecture and 3D-genome organization among different cell types, ranging from the size of the cell nucleus to higher-order chromosome structures and radial positioning of chromatin loci within the nucleus. These cell type-dependent variations in nuclear architecture and chromatin organization exhibit strong correlations with both the total transcriptional activity of the cell and transcriptional regulation of cell type-specific marker genes. Moreover, we found that the methylated DNA binding protein MeCP2 promotes active-inactive chromatin segregation and regulates transcription in a nuclear radial position-dependent manner that is highly correlated with its function in modulating active-inactive chromatin compartmentalization.

PMID:40009678 | DOI:10.1126/sciadv.adv2067