iScience. 2025 Jul 16;28(8):113121. doi: 10.1016/j.isci.2025.113121. eCollection 2025 Aug 15.

ABSTRACT

IRE1α and its downstream XBP1 signal is the most conserved unfolded protein response pathway that cells utilize to combat endoplasmic reticulum stress, also known to be utilized by tumor cells to adapt to harsh environment, leading to tumor progression. Several inhibitors against IRE1α have been developed, some of which show promising effect in clinical trial for cancer therapy, but none of them have been used in practice. Considering that hyper-activation of IRE1α induces cell death, we hypothesize that activation of IRE1α could be an alternative way for tumor suppression. Here, we identified divalent manganese ion as a potent activator to IRE1α, which interacts with the cytosolic part of IRE1α directly, augmenting the downstream pro-apoptotic pathway but not the pro-survival outcome. Mn2+ limits tumor growth in xenograft model in an IRE1α-dependent way. Our finding suggests pharmacological activation of IRE1α as an underestimated but promising way in cancer therapy.

PMID:40792035 | PMC:PMC12337650 | DOI:10.1016/j.isci.2025.113121

iScience. 2025 Jul 15;28(8):113112. doi: 10.1016/j.isci.2025.113112. eCollection 2025 Aug 15.

ABSTRACT

Prolactin is the main hormonal driver of mammalian lactation. To sustain milk production, basal prolactin levels must remain elevated compared to nonpregnant states. However, prolactin (23 kDa) is short-lived in circulation due to rapid renal excretion. Here, we design and test the galactopoietic effects of an engineered long-lasting prolactin in mice. The engineered variant, prolactin-extra long-acting (Prolactin-XL), is comprised of endogenously active human prolactin fused to an engineered human immunoglobulin G1 (IgG1) Fc domain. Prolactin-XL has a serum half-life of 70.9 h in mice, 2,625-fold longer than endogenously active human prolactin alone (70.9 h vs. 0.02 h). Prolactin-XL is engineered to be more susceptible to gastrointestinal proteases to reduce its uptake by nursing neonates. We demonstrate that Prolactin-XL increases lactation and restores growth of pups fed by dams with pharmacologically ablated lactation. We propose that Prolactin-XL is a potential tool for the study and pharmacologic stimulation of galactopoiesis.

PMID:40792031 | PMC:PMC12337787 | DOI:10.1016/j.isci.2025.113112

iScience. 2025 Jul 21;28(8):113165. doi: 10.1016/j.isci.2025.113165. eCollection 2025 Aug 15.

ABSTRACT

Colorectal cancer (CRC) is a prevalent digestive system malignancy accompanied by peritoneal metastasis occurring in 7% of cases. Methyltransferase-like 3 (METTL3) promoted the progression of CRC whereas its function in peritoneal metastasis was incompletely understood. Here, we found that METTL3 was upregulated in peritoneal metastasis tissues of CRC patients compared with CRC tissues. By sequencing the mRNA of above tissues, we discovered that METTL3-mediated N6-methyladenosine (m6A) modification regulated the downstream target neurexin-3 (NRXN3). NRXN3 promoted CRC peritoneal metastasis in vivo. Mechanically, we further verified the specific methylation sites of NRXN3 mRNA modified by METTL3. Functional cellular assays demonstrated METTL3-mediated upstream regulation of NRXN3. We also demonstrated that YTHDC1 is necessary for the METTL3-mediated stabilization of NRXN3 mRNA. Taken together, this study established a METTL3-YTHDC1-NRXN3 m6A modification-dependent regulation system in CRC peritoneal metastasis, suggesting potential candidates for peritoneal metastasis treatment.

PMID:40792021 | PMC:PMC12337687 | DOI:10.1016/j.isci.2025.113165

Biol Rev Camb Philos Soc. 2025 Aug 11. doi: 10.1111/brv.70064. Online ahead of print.

ABSTRACT

Theory predicts that facultatively asexual animals, which can leverage the advantages of both sexual and asexual reproduction, should outcompete obligately sexual and obligately asexual animals. Yet, paradoxically, obligate sexual reproduction predominates in many animal lineages, while the most flexible form of facultative asexuality (i.e. facultative parthenogenesis) appears to be rare. Recent theoretical work suggests that sexual conflict could help to resolve this paradox. Males that coercively fertilise females' eggs may, in the process, prevent alleles for parthenogenesis from spreading by limiting opportunities for asexual reproduction. Coercive males may also inhibit asexual reproduction by making resistance to sex disproportionately costly for females. In this review, we outline evidence of interactions with males that could impose costs on parthenogenetic females or hinder their ability to reproduce parthenogenetically in diverse animal taxa. The evidence suggests that such interactions between the sexes have the potential to mediate sexual conflict over mating and reproductive mode, both within facultative species and between closely related sexual and asexual taxa. However, the relative costs of sex and parthenogenesis are clearly context dependent, and much remains unknown. The most direct evidence for male inhibition of parthenogenesis comes from stick insects, but several other systems offer promising avenues for further investigation. Further research on the costs of mating and resistance in such systems could shed light on the reasons for the puzzling rarity of facultative parthenogenesis in nature.

PMID:40790891 | DOI:10.1111/brv.70064

Plant Genome. 2025 Sep;18(3):e70089. doi: 10.1002/tpg2.70089.

ABSTRACT

Stem rot, caused by Sclerotium rolfsii Sacc., is a devastating soil-borne disease causing up to 80% yield losses in groundnut globally. To dissect the genetic basis of resistance, we evaluated a diverse minicore germplasm panel over 3 years in stem rot sick-field conditions. Multi-locus genome-wide association study with the 58K single nucleotide polymorphisms (SNPs) Axiom_Arachis array genotyping identified 13 significant genomic regions associated with resistance across eight chromosomes with logarithm of the odds (LOD) scores ranging from 4.5 to 12.4 and R2 values between 6.9% and 58%. Within these regions, 145 candidate genes were implicated, including wall-associated receptor kinases, lucine-rich repeat and NB-ARC domain proteins, and peroxidase superfamily proteins. These genes orchestrate resistance through pathogen perception (e.g., receptor-like kinases), direct inhibition (R genes), toxin detoxification, and activation of transcription factors driving protective compound synthesis for cell recovery. If these defenses are compromised, a hypersensitive response-mediated apoptosis is triggered. Notably, resistance was exclusive to Virginia-type groundnut. The identified candidate genes showed strong correlation with RNA-seq data from stem rot-infected plants, reinforcing their role in the transcriptional defense response. Three kompetitive allele-specific PCR markers, namely, SnpAH00614 (on auxin-related gene AhSR001), SnpAH00625 (on histidine triad protein gene AhSR002), and SnpAH00626 (on E3 ubiquitin ligase gene AhSR003), were validated, confirming their significant contribution to stem rot resistance. These markers may facilitate the development of stem rot-resistant varieties through direct application in breeding programs through marker-assisted selection.

PMID:40790867 | DOI:10.1002/tpg2.70089

Food Res Int. 2025 Oct;218:116910. doi: 10.1016/j.foodres.2025.116910. Epub 2025 Jun 23.

ABSTRACT

Amylase/trypsin-inhibitors (ATIs) are known allergens and triggers of non-celiac wheat sensitivity. Until now, ATIs were only quantitated in wheat species. We developed and validated a targeted stable isotope dilution analysis LC-MS/MS method to quantitate ten barley-specific ATIs, including one monomeric and one dimeric amylase-inhibitor, four chloroform/methanol-soluble types, three subtilisin/chymotrypsin-inhibitors and one amylase/subtilisin-inhibitor. After successful validation in terms of precision, recovery and limits of detection and quantitation, the method was applied to 181 barley accessions from the Global EcoSeed panel, comprising 113 two-row and 68 six-row barleys of different genetic backgrounds. The overall ATI content was 1.1-5.2 mg/g, corresponding to 0.7-3.6 % of the total protein content with no clear distinction between two-row and six-row barleys. This study is the first to provide insights on the ATI content and composition of barley, which can be used to make low-ATI foods for special dietary needs.

PMID:40790697 | DOI:10.1016/j.foodres.2025.116910

BMC Plant Biol. 2025 Aug 11;25(1):1062. doi: 10.1186/s12870-025-07102-8.

ABSTRACT

Xylella fastidiosa, a highly pathogenic, xylem-limited, gram-negative bacterial species, represents a significant threat to many plant species, including olive, almond, grapevine, and alfalfa. Through cross-species transcriptomic analysis of Olea europaea, Prunus dulcis, Vitis vinifera, and Medicago sativa, we identified a novel core resistance network consisting of 18 conserved genes against Xylella fastidiosa, alongside 1852 divergent expression patterns. These common genes may play a crucial role in orchestrating a multi-layered plant defense response, enabling (1) structural reinforcement as well as facilitating cuticular wax biosynthesis (KCS11 and KAS1); (2) stress signaling mediated by hormonal crosstalk involving jasmonic acid (JA), salicylic acid (SA), and abscisic acid (ABA) mediated by the genes AOS and CYP707A4, alongside calcium signaling through ACA12 gene; (3) antimicrobial 22 compound production (β-amyrin synthase BAS, ABC transporter PDR6); and (4) resource optimization through trehalose metabolism (AT1G23870) and amino acid transport (AAP2). The protein-protein interaction networks revealed coordinated regulation of immune hubs including BAK1, WRKY33, and WRKY40, with novel connections to subtilase proteases and ubiquitin-proteasome components. This conserved molecular framework highlights evolutionary convergence in plant defenses against xylem pathogens, providing future targets for engineering resistance through cell wall modification, stress signaling potentiation, and secondary metabolite engineering.

PMID:40790398 | DOI:10.1186/s12870-025-07102-8

Nat Commun. 2025 Aug 11;16(1):7416. doi: 10.1038/s41467-025-62415-4.

ABSTRACT

Mathematical models are indispensable for studying the architecture and behavior of intracellular signaling networks. It is common to develop models using phenomenological approximations due to the difficulty of fully observing the intermediate steps in intracellular signaling pathways. Thus, multiple models can be built to represent the same pathway. This opens up challenges for model selection and decreases certainty in predictions. Here, we investigate Bayesian multimodel inference (MMI) as an approach to increase certainty in systems biology predictions, which becomes relevant when one wants to leverage a set of potentially incomplete models. Using existing models of the extracellular-regulated kinase (ERK) pathway, we show that MMI successfully combines models and yields predictors robust to model set changes and data uncertainties. We then use MMI to identify possible mechanisms of experimentally measured subcellular location-specific ERK activity. This work highlights MMI as a disciplined approach to increasing the certainty of intracellular signaling activity predictions.

PMID:40790297 | DOI:10.1038/s41467-025-62415-4

Front Pharmacol. 2025 Jul 28;16:1620732. doi: 10.3389/fphar.2025.1620732. eCollection 2025.

ABSTRACT

BACKGROUND: Curcumin, a key bioactive component of turmeric (Curcuma longa L. [Zingiberaceae]), has gained considerable attention for its potential to mitigate drug-induced toxicity. This review synthesizes and clarifies current findings on curcumin's ability to prevent the adverse effects of various pharmaceuticals.

METHODS: A comprehensive search using multiple databases-PubMed®, Scopus®, ScienceDirect®, and Web of Science®-was conducted for articles published up to October 2023. The current review is limited to randomized controlled trials, observational studies, and animal studies investigating the protective role of curcumin against drug-induced toxicity. The data extraction process included a variety of study characteristics, types of drugs used, curcumin dosing regimens, and reported outcomes associated with drug-induced toxicity.

RESULTS: A total of twenty-five studies were reviewed for this analysis. Curcumin may help reduce the side effects of certain medications, including sertraline, diclofenac, paclitaxel, irinotecan, and methotrexate.

DISCUSSION: Research also indicates that curcumin possesses antioxidant properties, reduces inflammation, and aids sperm production. Most importantly, sperm motility, density, and morphology significantly improved in curcumin-treated groups compared to the control groups undergoing toxic pharmaceutical treatment. The dosage of curcumin used in these studies ranges from 50 to 200 mg/kg body weight.

CONCLUSION: The available evidence suggests that curcumin may serve as a protective agent for male reproductive health against drug-induced damage, based on its diverse effects in mitigating oxidative stress and inflammation, which provide potential use in preserving reproductive health in males during pharmacological interventions. However, standardization of methodologies, along with more clinical evidence, is highly required before the practical application of findings related to treatment benefits can be made. Subsequent studies should focus on optimizing the use of this compound in combination with other pharmacological agents to enhance the protective effects of curcumin on male reproductive health.

PMID:40792206 | PMC:PMC12336216 | DOI:10.3389/fphar.2025.1620732

BMC Cancer. 2025 Aug 11;25(1):1303. doi: 10.1186/s12885-025-14733-5.

ABSTRACT

AIMS: The objective of this study was to analyze immune-related adverse events (irAEs) in a real-world data sample and examine the differences in incidence between affected organ systems, irAE severity, therapeutic agent, and gender.

METHODS: We retrospectively analyzed all consecutive patients treated with anti-cytotoxic T-lymphocyte associated protein 4 (CTLA-4) antibodies, anti-programmed death 1 (PD-1) inhibitors, and programmed death-ligand 1 (PD-L1) inhibitors between January 2020 and May 2023 in a tertiary referral center in Switzerland. IrAEs documented in the electronic health records (EHR) were graded according to the Common Terminology Criteria for Adverse Events (CTCAE) and analyzed descriptively.

RESULTS: Among the 500 patients, 196 (39.2%) were female. Treatments included pembrolizumab (51.2%), atezolizumab (20.2%), nivolumab (14.4%), durvalumab (6.4%), ipilimumab in combination with nivolumab (4.8%), cemiplimab (1.4%), avelumab (1.2%), and ipilimumab (0.4%). N = 216 (43.2%) patients had ≥ 1 irAEs (females: 47.4%; males: 40.5%). Severe (≥ grade 3) irAEs were reported in 13.6% of patients. The following irAE incidences were found: dermatological (15.2%), gastrointestinal (13.0%), endocrine (10.8%), musculoskeletal (4.8%), pulmonary (3.8%), systemic (3.6%), neurological (2.6%), cardiac (1.4%), renal (1.4%), hematological (0.6%), and ocular (0.2%).

CONCLUSION: Nearly half of the patients experienced ≥ 1 irAEs, of which one-third severe. Females experienced more irAEs than males, above all due to a higher incidence of grade 1 irAEs. Only about half of the irAEs were reported as coded diagnosis. Further prospective studies on irAEs are warranted using structured documentation.

PMID:40790180 | DOI:10.1186/s12885-025-14733-5

Sci Rep. 2025 Aug 11;15(1):29365. doi: 10.1038/s41598-025-09717-1.

ABSTRACT

Representation learning via pre-trained deep learning models is emerging as an integral method for studying the molecular structure-property relationship, which is then leveraged to predict molecular properties or design new molecules with desired attributes. We propose an unsupervised method to localize and characterize representations of pre-trained models through the lens of non-parametric property-driven subset scanning (PDSS), to improve the interpretability of deep molecular representations. We assess its detection capabilities on diverse molecular benchmarks (ZINC-250K, MOSES, MoleculeNet, FlavorDB, M2OR) across predictive chemical language models (MoLFormer, ChemBERTa) and molecular graph generative models (GraphAF, GCPN). We further study how representations evolve due to domain adaptation, and we evaluate the usefulness of the extracted property-driven elements in the embeddings as lower-dimension representations for downstream tasks. Experiments reveal notable information condensation in the pre-trained embeddings upon task-specific fine-tuning. For example, among the property-driven elements found in the embedding (out of [Formula: see text]), only 11 are shared between three distinct tasks (BACE, BBBP, and HIV), while [Formula: see text]-80 of those are unique to each task. Similar patterns are found for flavor and odor detection tasks. When we use the discovered property-driven elements as features for a new task, we find the same or improved performance (3 points up) while reducing the dimensions by 75% without fine-tuning required, thus indicating information localization.

PMID:40790135 | DOI:10.1038/s41598-025-09717-1

Radiother Oncol. 2025 Aug 9:111095. doi: 10.1016/j.radonc.2025.111095. Online ahead of print.

ABSTRACT

BACKGROUND AND PURPOSE: Accurate delineation of the Gross Tumor Volume (GTV) and the Internal Target Volume (ITV) in early-stage lung tumors is crucial in Stereotactic Body Radiation Therapy (SBRT). Traditionally, the ITVs, which account for breathing motion, are generated by manually contouring GTVs across all breathing phases (BPs), a time-consuming process. This research aims to streamline this workflow by developing a deep learning algorithm to automatically delineate GTVs in all four-dimensional computed tomography (4D-CT) BPs for early-stage Non-Small Cell Lung Cancer Patients (NSCLC).

METHODS: A dataset of 214 early-stage NSCLC patients treated with SBRT was used. Each patient had a 4D-CT scan containing ten reconstructed BPs. The data were divided into a training set (75 %) and a testing set (25 %). Three models SwinUNetR and Dynamic UNet (DynUnet), and a hybrid model combining both (Swin + Dyn)were trained and evaluated using the Dice Similarity Coefficient (DSC), 3 mm Surface Dice Similarity Coefficient (SDSC), and the 95th percentile Hausdorff distance (HD95). The best performing model was used to delineate GTVs in all test set BPs, creating the ITVs using two methods: all 10 phases and the maximum inspiration/expiration phases. The ITVs were compared to the ground truth ITVs.

RESULTS: The Swin + Dyn model achieved the highest performance, with a test set SDSC of 0.79 ± 0.14 for GTV 50 %. For the ITVs, the SDSC was 0.79 ± 0.16 using all 10 BPs and 0.77 ± 0.14 using 2 BPs. At the voxel level, the Swin + DynNet network achieved a sensitivity of 0.75 ± 0.14 and precision of 0.84 ± 0.10 for the ITV 2 breathing phases, and a sensitivity of 0.79 ± 0.12 and precision of 0.80 ± 0.11 for the 10 breathing phases.

CONCLUSION: The Swin + Dyn Net algorithm, trained on the maximum expiration CT-scan effectively delineated gross tumor volumes in all breathing phases and the resulting ITV showed a good agreement with the ground truth (surface DSC = 0.79 ± 0.16 using all 10 BPs and 0.77 ± 0.14 using 2 BPs.). The proposed approach could reduce delineation time and inter-performer variability in the tumor contouring process for NSCLC SBRT workflows.

PMID:40789427 | DOI:10.1016/j.radonc.2025.111095

Neuroscience. 2025 Aug 9:S0306-4522(25)00842-5. doi: 10.1016/j.neuroscience.2025.08.009. Online ahead of print.

ABSTRACT

The serotonin transporter (SERT) plays a key role in the regulation of serotonin levels in synapses and is significantly involved in the pathophysiology of major depressive disorder (MDD). In this review, molecular pathways connecting SERT dysfunction associated with depression are defined, including genetic polymorphisms (e.g., 5-HTTLPR), epigenetics (e.g., SLC6A4 methylation), and environmental interactions through stress and inflammation. Using a serotonergic pharmacophore, oxidative/nitrosative stress, cytokines, and neuroendocrine factors act via the hypothalamic-pituitary-adrenal (HPA) axis on SERT, whereas the crosstalk between SERT and brain-derived neurotrophic factor (BDNF) and cAMP response element-binding protein (CREB) signaling both bring about changes in mood and influence the response to treatment. Pharmacological treatments, such as SSRIs and SNRIs, aim at SERT; however, their effectiveness is limited due to interindividual variability. New treatments consist of allosteric modulators, multimodal antidepressants, and non-pharmacological treatments, involving exercise, diet, and microbiome manipulation. Tailor-made treatments involving the utilization of pharmacogenomics and neurobiological profiling can improve clinical outcomes. This review highlights SERT as a complex target in MDD and provides an argument in support of integrative, precision-focused interventions that aim to target the affective and cognitive symptoms of MDD.

PMID:40789515 | DOI:10.1016/j.neuroscience.2025.08.009

Sci Rep. 2025 Aug 11;15(1):29407. doi: 10.1038/s41598-025-15451-5.

ABSTRACT

Idiopathic osteosclerosis (IOS) and condensing osteitis (CO) represent radiopaque lesions often detected incidentally within the jaws, posing substantial diagnostic challenges due to their overlapping radiographic characteristics. The objective of this study was to assess the diagnostic efficacy of YOLOv8 and YOLOv11 deep learning algorithms in the identification of IOS and CO lesions on panoramic radiographs. A comprehensive collection of 1,000 panoramic images was retrospectively gathered and meticulously annotated utilizing a bounding box approach by two proficient oral and maxillofacial radiologists. All images were standardized to a resolution of 640 × 640 pixels and segregated into training (70%), validation (15%), and testing (15%) subsets. The performance of the models was evaluated based on metrics including accuracy, sensitivity, precision, F1 score, and the area under the receiver operating characteristic curve (AUC). YOLOv11 achieved notable precision scores of 98.8% for IOS and 97.1% for CO, alongside F1 scores of 96.8% and 95.6%, respectively. Conversely, YOLOv8 produced precision scores of 96.6% for IOS and 91.4% for CO, with F1 scores of 94% and 90%. These findings illustrate that AI-enhanced deep learning models possess the capability to accurately identify IOS and CO lesions, thereby presenting opportunities to improve diagnostic consistency, avert unnecessary invasive procedures, and facilitate more effective treatment planning within clinical practice.

PMID:40790082 | DOI:10.1038/s41598-025-15451-5

Sci Rep. 2025 Aug 12;15(1):29452. doi: 10.1038/s41598-025-14726-1.

ABSTRACT

Disease classification in maize plant is necessary for immediate treatment to enhance agricultural production and assure global food sustainability. Recent advancements in deep learning, specifically convolutional neural networks, have shown outstanding potential for image classification. This study presents Maize Net, a convolutional neural network model that precisely identifies diseases in maize leaves. Maize Net uses an attention mechanism to increase the model's efficiency by focusing on the relevant features and residual learning to improve the gradient flow. This also addresses the vanishing gradient problem while training deeper neural networks. A five-fold cross-validation test is conducted for generalization across the dataset, generating five models based on distinct training and testing sets. The macro-average of all evaluation metrics is considered to address the dataset's class imbalance problem. Maize Net achieved an average F1-score of 0.9509, recall of 0.9497, precision of 0.9525, and classification accuracy of 0.9595. These outcomes demonstrate MaizeNet's robustness and reliability in automated plant disease classification.

PMID:40790074 | DOI:10.1038/s41598-025-14726-1

Med Oncol. 2025 Aug 11;42(9):425. doi: 10.1007/s12032-025-02982-0.

ABSTRACT

Enhancing the accuracy of tumor response predictions enables the development of tailored therapeutic strategies for patients with breast cancer. In this study, we developed deep radiomic models to enhance the prediction of chemotherapy response after the first treatment cycle. 18F-Fludeoxyglucose PET/CT imaging data and clinical record from 60 breast cancer patients were retrospectively obtained from the Cancer Imaging Archive. PET/CT scans were conducted at three distinct stages of treatment; prior to the initiation of chemotherapy (T1), following the first cycle of chemotherapy (T2), and after the full chemotherapy regimen (T3). The patient's primary gross tumor volume (GTV) was delineated on PET images using a 40% threshold of the maximum standardized uptake value (SUVmax). Radiomic features were extracted from the GTV based on the PET/CT images. In addition, a squeeze-and-excitation network (SENet) deep learning model was employed to generate additional features from the PET/CT images for combined analysis. A XGBoost machine learning model was developed and compared with the conventional machine learning algorithm [random forest (RF), logistic regression (LR) and support vector machine (SVM)]. The performance of each model was assessed using receiver operating characteristics area under the curve (ROC AUC) analysis, and prediction accuracy in a validation cohort. Model performance was evaluated through fivefold cross-validation on the entire cohort, with data splits stratified by treatment response categories to ensure balanced representation. The AUC values for the machine learning models using only radiomic features were 0.85(XGBoost), 0.76 (RF), 0.80 (LR), and 0.59 (SVM), with XGBoost showing the best performance. After incorporating additional deep learning-derived features from SENet, the AUC values increased to 0.92, 0.88, 0.90, and 0.61, respectively, demonstrating significant improvements in predictive accuracy. Predictions were based on pre-treatment (T1) and post-first-cycle (T2) imaging data, enabling early assessment of chemotherapy response after the initial treatment cycle. Integrating deep learning-derived features significantly enhanced the performance of predictive models for chemotherapy response in breast cancer patients. This study demonstrated the superior predictive capability of the XGBoost model, emphasizing its potential to optimize personalized therapeutic strategies by accurately identifying patients unlikely to respond to chemotherapy after the first treatment cycle.

PMID:40790010 | DOI:10.1007/s12032-025-02982-0

Light Sci Appl. 2025 Aug 11;14(1):267. doi: 10.1038/s41377-025-01971-2.

ABSTRACT

We present a broadband and polarization-insensitive unidirectional imager that operates at the visible part of the spectrum, where image formation occurs in one direction, while in the opposite direction, it is blocked. This approach is enabled by deep learning-driven diffractive optical design with wafer-scale nano-fabrication using high-purity fused silica to ensure optical transparency and thermal stability. Our design achieves unidirectional imaging across three visible wavelengths (covering red, green, and blue parts of the spectrum), and we experimentally validated this broadband unidirectional imager by creating high-fidelity images in the forward direction and generating weak, distorted output patterns in the backward direction, in alignment with our numerical simulations. This work demonstrates wafer-scale production of diffractive optical processors, featuring 16 levels of nanoscale phase features distributed across two axially aligned diffractive layers for visible unidirectional imaging. This approach facilitates mass-scale production of ~0.5 billion nanoscale phase features per wafer, supporting high-throughput manufacturing of hundreds to thousands of multi-layer diffractive processors suitable for large apertures and parallel processing of multiple tasks. Beyond broadband unidirectional imaging in the visible spectrum, this study establishes a pathway for artificial-intelligence-enabled diffractive optics with versatile applications, signaling a new era in optical device functionality with industrial-level, massively scalable fabrication.

PMID:40789836 | DOI:10.1038/s41377-025-01971-2

Annu Rev Biomed Data Sci. 2025 Aug;8(1):149-171. doi: 10.1146/annurev-biodatasci-103123-095814.

ABSTRACT

The rapid development of artificial intelligence (AI) has had a significant impact on medical research, introducing new possibilities for pathology studies. There is a recent trend of applying large-scale AI models to many fields, and this trend has given rise to the pathology foundation models and pathology ensemble models. Large models in pathology are not standalone innovations; they build upon a legacy where AI has consistently played a vital role in pathology studies long before their advent. Numerous pathology datasets and AI models have been developed to support advancements in the field, with these combined efforts paving the way for the emergence of large models in pathology. AI greatly enhances pathology studies, yet its widespread use in sensitive applications also raises significant ethical concerns, including privacy risks. In this review, we summarize the datasets and models that are useful to pathology studies, with a particular focus on how they illuminate the path toward large-scale applications.

PMID:40789736 | DOI:10.1146/annurev-biodatasci-103123-095814

Annu Rev Biomed Data Sci. 2025 Aug;8(1):101-124. doi: 10.1146/annurev-biodatasci-103023-050856.

ABSTRACT

Deep learning and artificial intelligence (AI) have seen explosive growth and success in biomedical applications in the last decade, largely due to the rapid development of deep neural networks and their underlying neural network (NN) architectures. Here, we explore biomedical deep learning and AI from the specific perspective of NN architectures. We discuss widely varying design principles of NN architectures, their use in particular biomedical applications, and the assumptions (often hidden) built into them. We explore neural architecture search techniques that automate the design of NN topology to optimize task performance. Advanced neural architectures are being developed for both molecular and healthcare applications, employing elements of graph networks, transformers, and interpretable NNs, and we discuss and summarize the design considerations and unique advantages of each architecture. Future advances will include the employment of multimodal language models and smaller highly focused mechanistic models that build on the success of today's large models.

PMID:40789735 | DOI:10.1146/annurev-biodatasci-103023-050856

SLAS Technol. 2025 Aug 9:100338. doi: 10.1016/j.slast.2025.100338. Online ahead of print.

ABSTRACT

Optic disc and cup segmentation is a crucial subfield of computer vision, playing a pivotal role in automated pathological image analysis. It enables precise, efficient, and automated diagnosis of ocular conditions, significantly aiding clinicians in real-world medical applications. However, due to the scarcity of medical segmentation data and the insufficient integration of global contextual information, the segmentation accuracy remains suboptimal. This issue becomes particularly pronounced in optic disc and cup cases with complex anatomical structures and ambiguous boundaries.In order to address these limitations, this paper introduces a self-supervised training strategy integrated with a newly designed network architecture to improve segmentation accuracy.Specifically,we initially propose a non-local dual deformable convolutional block,which aims to capture the irregular image patterns(i.e. boundary).Secondly,we modify the traditional vision transformer and design an adaptive K-Nearest Neighbors(KNN) transformation block to extract the global semantic context from images. Finally,an initialization strategy based on self-supervised training is proposed to reduce the burden on the network on labeled data.Comprehensive experimental evaluations demonstrate the effectiveness of our proposed method, which outperforms previous networks and achieves state-of-the-art performance,with IOU scores of 0.9577 for the optic disc and 0.8399 for the optic cup on the REFUGE dataset.

PMID:40789537 | DOI:10.1016/j.slast.2025.100338