Only content delivered through the publication boundary on this date is included.
Terence Tao blog RSS In this blog post, Terence Tao announces that the SAIR Foundation is launching an Open Math Model initiative, inviting the mathematical community and its supporters to build open-source models together and openly seeking partners who can contribute funding, compute, expertise, or community building; the initiative sets out principles of community-shaped models, tools for everyday mathematical work (understanding difficult arguments, checking references, exploring examples, writing code, and formalizing proofs), open development (open licensed weights and code, published training methods, reproducible evaluations), community ownership of data, shared intellectual property (for example Apache 2.0, MIT, or CC BY 4.
In this blog post, Terence Tao announces that the SAIR Foundation is launching an Open Math Model initiative, inviting the mathematical community and its supporters to build open-source models together and openly seeking partners who can contribute funding, compute, expertise, or community building; the initiative sets out principles of community-shaped models, tools for everyday mathematical work (understanding difficult arguments, checking references, exploring examples, writing code, and formalizing proofs), open development (open licensed weights and code, published training methods, reproducible evaluations), community ownership of data, shared intellectual property (for example Apache 2.0, MIT, or CC BY 4.
In this blog post, Terence Tao announces that the SAIR Foundation is launching an Open Math Model initiative, inviting the mathematical community and its supporters to build open-source models together and openly seeking partners who can contribute funding, compute, expertise, or community building; the initiative sets out principles of community-shaped models, tools for everyday mathematical work (understanding difficult arguments, checking references, exploring examples, writing code, and formalizing proofs), open development (open licensed weights and code, published training methods, reproducible evaluations), community ownership of data, shared intellectual property (for example Apache 2.0, MIT, or CC BY 4.
In this blog post, Terence Tao announces that the SAIR Foundation is launching an Open Math Model initiative, inviting the mathematical community and its supporters to build open-source models together and openly seeking partners who can contribute funding, compute, expertise, or community building; the initiative sets out principles of community-shaped models, tools for everyday mathematical work (understanding difficult arguments, checking references, exploring examples, writing code, and formalizing proofs), open development (open licensed weights and code, published training methods, reproducible evaluations), community ownership of data, shared intellectual property (for example Apache 2.0, MIT, or CC BY 4.
MIT Technology Review After a subscriber Roundtables event, MIT Technology Review's senior AI editor Will Douglas Heaven and AI reporter Grace Huckins answer reader questions about existential risk from AI, offering reporter-level judgments on personal risk, why AI might cause harm, the state of alignment research, corporate motives, the autonomy-control trade-off, regulation, and whether the discussion itself could become self-fulfilling.
After a subscriber Roundtables event, MIT Technology Review's senior AI editor Will Douglas Heaven and AI reporter Grace Huckins answer reader questions about existential risk from AI, offering reporter-level judgments on personal risk, why AI might cause harm, the state of alignment research, corporate motives, the autonomy-control trade-off, regulation, and whether the discussion itself could become self-fulfilling.
After a subscriber Roundtables event, MIT Technology Review's senior AI editor Will Douglas Heaven and AI reporter Grace Huckins answer reader questions about existential risk from AI, offering reporter-level judgments on personal risk, why AI might cause harm, the state of alignment research, corporate motives, the autonomy-control trade-off, regulation, and whether the discussion itself could become self-fulfilling.
After a subscriber Roundtables event, MIT Technology Review's senior AI editor Will Douglas Heaven and AI reporter Grace Huckins answer reader questions about existential risk from AI, offering reporter-level judgments on personal risk, why AI might cause harm, the state of alignment research, corporate motives, the autonomy-control trade-off, regulation, and whether the discussion itself could become self-fulfilling.
MIT Technology Review This MIT Technology Review article surveys concerns voiced by AI company leaders and researchers that AI could aid the design, creation, and release of bioweapons, citing the 2022 case in which Collaborations Pharmaceuticals' AI "molecule generator" produced 40,000 molecules with potential as chemical warfare agents in under six hours and Anthropic's report acknowledging attempts to use its models to explore making chikungunya virus more transmissible and creating a more dangerous bird flu, while also presenting dissenting views from some Imperial College London biologists that AI tools are not yet good enough to fully develop bioweapons and that the greatest pandemic risk comes from already-circulating pathogens such as H5N1, and summarizing existing safeguards—DNA order screening, red-te
This MIT Technology Review article surveys concerns voiced by AI company leaders and researchers that AI could aid the design, creation, and release of bioweapons, citing the 2022 case in which Collaborations Pharmaceuticals' AI "molecule generator" produced 40,000 molecules with potential as chemical warfare agents in under six hours and Anthropic's report acknowledging attempts to use its models to explore making chikungunya virus more transmissible and creating a more dangerous bird flu, while also presenting dissenting views from some Imperial College London biologists that AI tools are not yet good enough to fully develop bioweapons and that the greatest pandemic risk comes from already-circulating pathogens such as H5N1, and summarizing existing safeguards—DNA order screening, red-te
This MIT Technology Review article surveys concerns voiced by AI company leaders and researchers that AI could aid the design, creation, and release of bioweapons, citing the 2022 case in which Collaborations Pharmaceuticals' AI "molecule generator" produced 40,000 molecules with potential as chemical warfare agents in under six hours and Anthropic's report acknowledging attempts to use its models to explore making chikungunya virus more transmissible and creating a more dangerous bird flu, while also presenting dissenting views from some Imperial College London biologists that AI tools are not yet good enough to fully develop bioweapons and that the greatest pandemic risk comes from already-circulating pathogens such as H5N1, and summarizing existing safeguards—DNA order screening, red-te
This MIT Technology Review article surveys concerns voiced by AI company leaders and researchers that AI could aid the design, creation, and release of bioweapons, citing the 2022 case in which Collaborations Pharmaceuticals' AI "molecule generator" produced 40,000 molecules with potential as chemical warfare agents in under six hours and Anthropic's report acknowledging attempts to use its models to explore making chikungunya virus more transmissible and creating a more dangerous bird flu, while also presenting dissenting views from some Imperial College London biologists that AI tools are not yet good enough to fully develop bioweapons and that the greatest pandemic risk comes from already-circulating pathogens such as H5N1, and summarizing existing safeguards—DNA order screening, red-te
bioRxiv The study proposes RSAUNet, a deep learning architecture that combines residual convolutional blocks, Swin Transformer blocks, and attention mechanisms inside a U-Net backbone for MRI prostate cancer segmentation, reporting a Dice coefficient of 0.998 and a Jaccard index (IoU) of 0.965 on a public Kaggle prostate annotation dataset, together with an ablation study tracking loss, accuracy, Dice, and mean IoU as each component is added.
The study proposes RSAUNet, a deep learning architecture that combines residual convolutional blocks, Swin Transformer blocks, and attention mechanisms inside a U-Net backbone for MRI prostate cancer segmentation, reporting a Dice coefficient of 0.998 and a Jaccard index (IoU) of 0.965 on a public Kaggle prostate annotation dataset, together with an ablation study tracking loss, accuracy, Dice, and mean IoU as each component is added.
The study proposes RSAUNet, a deep learning architecture that combines residual convolutional blocks, Swin Transformer blocks, and attention mechanisms inside a U-Net backbone for MRI prostate cancer segmentation, reporting a Dice coefficient of 0.998 and a Jaccard index (IoU) of 0.965 on a public Kaggle prostate annotation dataset, together with an ablation study tracking loss, accuracy, Dice, and mean IoU as each component is added.
The study proposes RSAUNet, a deep learning architecture that combines residual convolutional blocks, Swin Transformer blocks, and attention mechanisms inside a U-Net backbone for MRI prostate cancer segmentation, reporting a Dice coefficient of 0.998 and a Jaccard index (IoU) of 0.965 on a public Kaggle prostate annotation dataset, together with an ablation study tracking loss, accuracy, Dice, and mean IoU as each component is added.
bioRxiv This work releases curated 2020-2021 Genomes to Fields imagery from 356 drone flights across 19 environments covering 1,180 maize hybrids, and uses functional principal components of vegetation index and weather trajectories together with genomic information for yield prediction and QTL mapping, finding that combined genomic and phenomic kernels raised held-out hybrid prediction to r = 0.501 when environments were represented in training and 0.408 when environments were also withheld, that accumulated growing degree days offered no consistent advantage over days after planting, that weather contributed modest task-dependent gains, and that NGRDI functional principal components repeatedly mapped to quantitative trait loci on chromosomes 3 and 7.
This work releases curated 2020-2021 Genomes to Fields imagery from 356 drone flights across 19 environments covering 1,180 maize hybrids, and uses functional principal components of vegetation index and weather trajectories together with genomic information for yield prediction and QTL mapping, finding that combined genomic and phenomic kernels raised held-out hybrid prediction to r = 0.501 when environments were represented in training and 0.408 when environments were also withheld, that accumulated growing degree days offered no consistent advantage over days after planting, that weather contributed modest task-dependent gains, and that NGRDI functional principal components repeatedly mapped to quantitative trait loci on chromosomes 3 and 7.
This work releases curated 2020-2021 Genomes to Fields imagery from 356 drone flights across 19 environments covering 1,180 maize hybrids, and uses functional principal components of vegetation index and weather trajectories together with genomic information for yield prediction and QTL mapping, finding that combined genomic and phenomic kernels raised held-out hybrid prediction to r = 0.501 when environments were represented in training and 0.408 when environments were also withheld, that accumulated growing degree days offered no consistent advantage over days after planting, that weather contributed modest task-dependent gains, and that NGRDI functional principal components repeatedly mapped to quantitative trait loci on chromosomes 3 and 7.
This work releases curated 2020-2021 Genomes to Fields imagery from 356 drone flights across 19 environments covering 1,180 maize hybrids, and uses functional principal components of vegetation index and weather trajectories together with genomic information for yield prediction and QTL mapping, finding that combined genomic and phenomic kernels raised held-out hybrid prediction to r = 0.501 when environments were represented in training and 0.408 when environments were also withheld, that accumulated growing degree days offered no consistent advantage over days after planting, that weather contributed modest task-dependent gains, and that NGRDI functional principal components repeatedly mapped to quantitative trait loci on chromosomes 3 and 7.
Current opinion in oncology This review indicates that TROP2-directed antibody-drug conjugates (ADCs) are becoming increasingly integrated into breast cancer treatment across multiple disease settings, and that although sacituzumab govitecan and datopotamab deruxtecan share TROP2 targeting and topoisomerase I inhibition, their safety profiles differ—sacituzumab govitecan is mainly associated with neutropenia and diarrhoea, whereas datopotamab deruxtecan is characterized by stomatitis, ocular surface events and a low but clinically relevant risk of interstitial lung disease/pneumonitis—differences attributed to the integrated effects of payload, linker stability, drug-to-antibody ratio, tissue distribution and target-independent uptake, with corresponding agent-specific, proactive and increasingly individualized preve
This review indicates that TROP2-directed antibody-drug conjugates (ADCs) are becoming increasingly integrated into breast cancer treatment across multiple disease settings, and that although sacituzumab govitecan and datopotamab deruxtecan share TROP2 targeting and topoisomerase I inhibition, their safety profiles differ—sacituzumab govitecan is mainly associated with neutropenia and diarrhoea, whereas datopotamab deruxtecan is characterized by stomatitis, ocular surface events and a low but clinically relevant risk of interstitial lung disease/pneumonitis—differences attributed to the integrated effects of payload, linker stability, drug-to-antibody ratio, tissue distribution and target-independent uptake, with corresponding agent-specific, proactive and increasingly individualized preve
This review indicates that TROP2-directed antibody-drug conjugates (ADCs) are becoming increasingly integrated into breast cancer treatment across multiple disease settings, and that although sacituzumab govitecan and datopotamab deruxtecan share TROP2 targeting and topoisomerase I inhibition, their safety profiles differ—sacituzumab govitecan is mainly associated with neutropenia and diarrhoea, whereas datopotamab deruxtecan is characterized by stomatitis, ocular surface events and a low but clinically relevant risk of interstitial lung disease/pneumonitis—differences attributed to the integrated effects of payload, linker stability, drug-to-antibody ratio, tissue distribution and target-independent uptake, with corresponding agent-specific, proactive and increasingly individualized preve
This review indicates that TROP2-directed antibody-drug conjugates (ADCs) are becoming increasingly integrated into breast cancer treatment across multiple disease settings, and that although sacituzumab govitecan and datopotamab deruxtecan share TROP2 targeting and topoisomerase I inhibition, their safety profiles differ—sacituzumab govitecan is mainly associated with neutropenia and diarrhoea, whereas datopotamab deruxtecan is characterized by stomatitis, ocular surface events and a low but clinically relevant risk of interstitial lung disease/pneumonitis—differences attributed to the integrated effects of payload, linker stability, drug-to-antibody ratio, tissue distribution and target-independent uptake, with corresponding agent-specific, proactive and increasingly individualized preve
bioRxiv The work introduces LDDM (Large Drug Discovery Model), a unified 3D generative framework supporting constrained and unconstrained docking, fragment linking and growing, and de novo design, together with a programmable design algorithm that produces synthetically accessible compounds satisfying fine-grained objectives; the authors experimentally validated designed or optimised ligands for five therapeutically relevant protein targets, achieving high success rates and identifying molecules with confirmed binding affinity while synthesizing only a small number of generated compounds, with the best designs structurally characterised by NMR spectroscopy and X-ray crystallography indicating high prediction accuracy.
The work introduces LDDM (Large Drug Discovery Model), a unified 3D generative framework supporting constrained and unconstrained docking, fragment linking and growing, and de novo design, together with a programmable design algorithm that produces synthetically accessible compounds satisfying fine-grained objectives; the authors experimentally validated designed or optimised ligands for five therapeutically relevant protein targets, achieving high success rates and identifying molecules with confirmed binding affinity while synthesizing only a small number of generated compounds, with the best designs structurally characterised by NMR spectroscopy and X-ray crystallography indicating high prediction accuracy.
The work introduces LDDM (Large Drug Discovery Model), a unified 3D generative framework supporting constrained and unconstrained docking, fragment linking and growing, and de novo design, together with a programmable design algorithm that produces synthetically accessible compounds satisfying fine-grained objectives; the authors experimentally validated designed or optimised ligands for five therapeutically relevant protein targets, achieving high success rates and identifying molecules with confirmed binding affinity while synthesizing only a small number of generated compounds, with the best designs structurally characterised by NMR spectroscopy and X-ray crystallography indicating high prediction accuracy.
The work introduces LDDM (Large Drug Discovery Model), a unified 3D generative framework supporting constrained and unconstrained docking, fragment linking and growing, and de novo design, together with a programmable design algorithm that produces synthetically accessible compounds satisfying fine-grained objectives; the authors experimentally validated designed or optimised ligands for five therapeutically relevant protein targets, achieving high success rates and identifying molecules with confirmed binding affinity while synthesizing only a small number of generated compounds, with the best designs structurally characterised by NMR spectroscopy and X-ray crystallography indicating high prediction accuracy.
bioRxiv The study introduces OpenLipid, a large language model (LLM)-based workflow for targeted DIA lipidomics that builds assay libraries from DDA results and, in a zero-shot setting, directly evaluates extracted ion chromatograms (XICs) from DIA data to select target lipid peaks and produce human-readable rationales; across four human plasma and mouse feces datasets in positive and negative ionization modes it identified 55.3%/19.0% (plasma) and 71.8%/31.8% (feces) of library targets at 5% FDR, comparable overall to DIAMetAlyzer (57.8%/19.7%, 89.0%/17.8%) and substantially higher than untargeted MS-DIAL DIA (12.6%/0.0%, 47.5%/0.
The study introduces OpenLipid, a large language model (LLM)-based workflow for targeted DIA lipidomics that builds assay libraries from DDA results and, in a zero-shot setting, directly evaluates extracted ion chromatograms (XICs) from DIA data to select target lipid peaks and produce human-readable rationales; across four human plasma and mouse feces datasets in positive and negative ionization modes it identified 55.3%/19.0% (plasma) and 71.8%/31.8% (feces) of library targets at 5% FDR, comparable overall to DIAMetAlyzer (57.8%/19.7%, 89.0%/17.8%) and substantially higher than untargeted MS-DIAL DIA (12.6%/0.0%, 47.5%/0.
The study introduces OpenLipid, a large language model (LLM)-based workflow for targeted DIA lipidomics that builds assay libraries from DDA results and, in a zero-shot setting, directly evaluates extracted ion chromatograms (XICs) from DIA data to select target lipid peaks and produce human-readable rationales; across four human plasma and mouse feces datasets in positive and negative ionization modes it identified 55.3%/19.0% (plasma) and 71.8%/31.8% (feces) of library targets at 5% FDR, comparable overall to DIAMetAlyzer (57.8%/19.7%, 89.0%/17.8%) and substantially higher than untargeted MS-DIAL DIA (12.6%/0.0%, 47.5%/0.
The study introduces OpenLipid, a large language model (LLM)-based workflow for targeted DIA lipidomics that builds assay libraries from DDA results and, in a zero-shot setting, directly evaluates extracted ion chromatograms (XICs) from DIA data to select target lipid peaks and produce human-readable rationales; across four human plasma and mouse feces datasets in positive and negative ionization modes it identified 55.3%/19.0% (plasma) and 71.8%/31.8% (feces) of library targets at 5% FDR, comparable overall to DIAMetAlyzer (57.8%/19.7%, 89.0%/17.8%) and substantially higher than untargeted MS-DIAL DIA (12.6%/0.0%, 47.5%/0.
World journal of transplantation This review summarizes applications of artificial intelligence and machine learning across the transplantation surgery care pathway, covering preoperative anatomical assessment, graft weight estimation and candidate selection, perioperative prediction of massive transfusion, hemorrhage and acute kidney injury plus robotic-assisted surgery, postoperative early prediction of sepsis, pneumonia and graft dysfunction with long-term monitoring, and cross-cutting innovations such as hyperspectral imaging and automated histopathological evaluation, while noting that multimodal models integrating electronic health records, intraoperative signals, ultrasound and histology can bridge diagnostic, prognostic and therapeutic decision-making, though clinical translation still requires rigorous validation
This review summarizes applications of artificial intelligence and machine learning across the transplantation surgery care pathway, covering preoperative anatomical assessment, graft weight estimation and candidate selection, perioperative prediction of massive transfusion, hemorrhage and acute kidney injury plus robotic-assisted surgery, postoperative early prediction of sepsis, pneumonia and graft dysfunction with long-term monitoring, and cross-cutting innovations such as hyperspectral imaging and automated histopathological evaluation, while noting that multimodal models integrating electronic health records, intraoperative signals, ultrasound and histology can bridge diagnostic, prognostic and therapeutic decision-making, though clinical translation still requires rigorous validation
This review summarizes applications of artificial intelligence and machine learning across the transplantation surgery care pathway, covering preoperative anatomical assessment, graft weight estimation and candidate selection, perioperative prediction of massive transfusion, hemorrhage and acute kidney injury plus robotic-assisted surgery, postoperative early prediction of sepsis, pneumonia and graft dysfunction with long-term monitoring, and cross-cutting innovations such as hyperspectral imaging and automated histopathological evaluation, while noting that multimodal models integrating electronic health records, intraoperative signals, ultrasound and histology can bridge diagnostic, prognostic and therapeutic decision-making, though clinical translation still requires rigorous validation
This review summarizes applications of artificial intelligence and machine learning across the transplantation surgery care pathway, covering preoperative anatomical assessment, graft weight estimation and candidate selection, perioperative prediction of massive transfusion, hemorrhage and acute kidney injury plus robotic-assisted surgery, postoperative early prediction of sepsis, pneumonia and graft dysfunction with long-term monitoring, and cross-cutting innovations such as hyperspectral imaging and automated histopathological evaluation, while noting that multimodal models integrating electronic health records, intraoperative signals, ultrasound and histology can bridge diagnostic, prognostic and therapeutic decision-making, though clinical translation still requires rigorous validation
Journal of the American Heart Association This retrospective, single-center external validation study enrolled 100 consecutive patients undergoing clinically indicated OCT and used the previously developed OCT-AID algorithm to perform automated pixelwise full-vessel labeling of 2560 analyzable frames, comparing it frame by frame against an expert manual reference standard; agreement was excellent for calcified plaque identification (κ=0.88) and quantification (intraclass correlation coefficients 0.79–0.93), close to interobserver variability, reasonable for lipid plaque identification and quantification (κ=0.68; lipid arc intraclass correlation coefficient 0.79; minimum fibrous cap thickness intraclass correlation coefficient 0.
This retrospective, single-center external validation study enrolled 100 consecutive patients undergoing clinically indicated OCT and used the previously developed OCT-AID algorithm to perform automated pixelwise full-vessel labeling of 2560 analyzable frames, comparing it frame by frame against an expert manual reference standard; agreement was excellent for calcified plaque identification (κ=0.88) and quantification (intraclass correlation coefficients 0.79–0.93), close to interobserver variability, reasonable for lipid plaque identification and quantification (κ=0.68; lipid arc intraclass correlation coefficient 0.79; minimum fibrous cap thickness intraclass correlation coefficient 0.
This retrospective, single-center external validation study enrolled 100 consecutive patients undergoing clinically indicated OCT and used the previously developed OCT-AID algorithm to perform automated pixelwise full-vessel labeling of 2560 analyzable frames, comparing it frame by frame against an expert manual reference standard; agreement was excellent for calcified plaque identification (κ=0.88) and quantification (intraclass correlation coefficients 0.79–0.93), close to interobserver variability, reasonable for lipid plaque identification and quantification (κ=0.68; lipid arc intraclass correlation coefficient 0.79; minimum fibrous cap thickness intraclass correlation coefficient 0.
This retrospective, single-center external validation study enrolled 100 consecutive patients undergoing clinically indicated OCT and used the previously developed OCT-AID algorithm to perform automated pixelwise full-vessel labeling of 2560 analyzable frames, comparing it frame by frame against an expert manual reference standard; agreement was excellent for calcified plaque identification (κ=0.88) and quantification (intraclass correlation coefficients 0.79–0.93), close to interobserver variability, reasonable for lipid plaque identification and quantification (κ=0.68; lipid arc intraclass correlation coefficient 0.79; minimum fibrous cap thickness intraclass correlation coefficient 0.
bioRxiv This work presents LLMsFold, a computational framework that identifies binding pockets geometrically, has Llama-3.3-70B generate candidate small molecules as SMILES strings, evaluates each with the Boltz-2 co-folding model for bound pose and binding affinity, and iteratively refines candidates through a feedback loop, yielding molecules for ACVR1 and CD19 that pass drug-likeness, synthetic accessibility, and novelty filters, with the ACVR1 candidate reaching a predicted affinity probability of 0.953 and predicted pIC50 of about 10.72 and the CD19 Pocket 1 candidate reaching a predicted pIC50 of about 7.73.
This work presents LLMsFold, a computational framework that identifies binding pockets geometrically, has Llama-3.3-70B generate candidate small molecules as SMILES strings, evaluates each with the Boltz-2 co-folding model for bound pose and binding affinity, and iteratively refines candidates through a feedback loop, yielding molecules for ACVR1 and CD19 that pass drug-likeness, synthetic accessibility, and novelty filters, with the ACVR1 candidate reaching a predicted affinity probability of 0.953 and predicted pIC50 of about 10.72 and the CD19 Pocket 1 candidate reaching a predicted pIC50 of about 7.73.
This work presents LLMsFold, a computational framework that identifies binding pockets geometrically, has Llama-3.3-70B generate candidate small molecules as SMILES strings, evaluates each with the Boltz-2 co-folding model for bound pose and binding affinity, and iteratively refines candidates through a feedback loop, yielding molecules for ACVR1 and CD19 that pass drug-likeness, synthetic accessibility, and novelty filters, with the ACVR1 candidate reaching a predicted affinity probability of 0.953 and predicted pIC50 of about 10.72 and the CD19 Pocket 1 candidate reaching a predicted pIC50 of about 7.73.
This work presents LLMsFold, a computational framework that identifies binding pockets geometrically, has Llama-3.3-70B generate candidate small molecules as SMILES strings, evaluates each with the Boltz-2 co-folding model for bound pose and binding affinity, and iteratively refines candidates through a feedback loop, yielding molecules for ACVR1 and CD19 that pass drug-likeness, synthetic accessibility, and novelty filters, with the ACVR1 candidate reaching a predicted affinity probability of 0.953 and predicted pIC50 of about 10.72 and the CD19 Pocket 1 candidate reaching a predicted pIC50 of about 7.73.
bioRxiv This work introduces latent generative search for binder design, a framework that uses reward-guided search at inference time to steer the Proteina-Complexa generative model, which codesigns sequence and structure together in a continuous latent space and thereby removes the inverse-folding step; in a screen of more than one million designs by multiplexed phage display, it produced more validated binders than every other method tested, its codesigned sequences surpassed post hoc redesign, it delivered high-affinity binders across therapeutic receptors, a viral attachment protein and intracellular signalling targets, and it generated the first de novo proteins that bind a free carbohydrate, including one that discriminates between blood-group antigens.
This work introduces latent generative search for binder design, a framework that uses reward-guided search at inference time to steer the Proteina-Complexa generative model, which codesigns sequence and structure together in a continuous latent space and thereby removes the inverse-folding step; in a screen of more than one million designs by multiplexed phage display, it produced more validated binders than every other method tested, its codesigned sequences surpassed post hoc redesign, it delivered high-affinity binders across therapeutic receptors, a viral attachment protein and intracellular signalling targets, and it generated the first de novo proteins that bind a free carbohydrate, including one that discriminates between blood-group antigens.
This work introduces latent generative search for binder design, a framework that uses reward-guided search at inference time to steer the Proteina-Complexa generative model, which codesigns sequence and structure together in a continuous latent space and thereby removes the inverse-folding step; in a screen of more than one million designs by multiplexed phage display, it produced more validated binders than every other method tested, its codesigned sequences surpassed post hoc redesign, it delivered high-affinity binders across therapeutic receptors, a viral attachment protein and intracellular signalling targets, and it generated the first de novo proteins that bind a free carbohydrate, including one that discriminates between blood-group antigens.
This work introduces latent generative search for binder design, a framework that uses reward-guided search at inference time to steer the Proteina-Complexa generative model, which codesigns sequence and structure together in a continuous latent space and thereby removes the inverse-folding step; in a screen of more than one million designs by multiplexed phage display, it produced more validated binders than every other method tested, its codesigned sequences surpassed post hoc redesign, it delivered high-affinity binders across therapeutic receptors, a viral attachment protein and intracellular signalling targets, and it generated the first de novo proteins that bind a free carbohydrate, including one that discriminates between blood-group antigens.
Journal of global health This study compiled a national monthly series of hepatitis B notifications in mainland China from January 2004 to December 2025, applied CEEMDAN to isolate multiscale temporal components, and trained four models (GRU, CNN, SVM, and a Transformer encoder) with KOA-optimised hyperparameters on a sliding 12-month window recursively extended to 24-month horizons; the Transformer delivered the best out-of-sample fit on the held-out test split (MAE = 3105.508, MAPE = 0.024, RMSE = 4071.901, R² = 0.928), SVM ranked second, CNN outperformed GRU but not SVM, and forecasts for 2026-2027 remain elevated, signalling little improvement and even possible resurgence.
This study compiled a national monthly series of hepatitis B notifications in mainland China from January 2004 to December 2025, applied CEEMDAN to isolate multiscale temporal components, and trained four models (GRU, CNN, SVM, and a Transformer encoder) with KOA-optimised hyperparameters on a sliding 12-month window recursively extended to 24-month horizons; the Transformer delivered the best out-of-sample fit on the held-out test split (MAE = 3105.508, MAPE = 0.024, RMSE = 4071.901, R² = 0.928), SVM ranked second, CNN outperformed GRU but not SVM, and forecasts for 2026-2027 remain elevated, signalling little improvement and even possible resurgence.
This study compiled a national monthly series of hepatitis B notifications in mainland China from January 2004 to December 2025, applied CEEMDAN to isolate multiscale temporal components, and trained four models (GRU, CNN, SVM, and a Transformer encoder) with KOA-optimised hyperparameters on a sliding 12-month window recursively extended to 24-month horizons; the Transformer delivered the best out-of-sample fit on the held-out test split (MAE = 3105.508, MAPE = 0.024, RMSE = 4071.901, R² = 0.928), SVM ranked second, CNN outperformed GRU but not SVM, and forecasts for 2026-2027 remain elevated, signalling little improvement and even possible resurgence.
This study compiled a national monthly series of hepatitis B notifications in mainland China from January 2004 to December 2025, applied CEEMDAN to isolate multiscale temporal components, and trained four models (GRU, CNN, SVM, and a Transformer encoder) with KOA-optimised hyperparameters on a sliding 12-month window recursively extended to 24-month horizons; the Transformer delivered the best out-of-sample fit on the held-out test split (MAE = 3105.508, MAPE = 0.024, RMSE = 4071.901, R² = 0.928), SVM ranked second, CNN outperformed GRU but not SVM, and forecasts for 2026-2027 remain elevated, signalling little improvement and even possible resurgence.
Current opinion in oncology This review focuses on studies published over the past 18 months and examines the role of molecular PET imaging in differentiating treatment-related changes, especially pseudoprogression, from true tumor progression in gliomas, suggesting that static amino acid PET, particularly [18F]FET, may provide diagnostic information comparable to dynamic acquisition while simplifying protocols, that [68Ga]FAPI PET may add information on the tumor microenvironment, and that quantitative PET parameters are affected by reconstruction algorithms, reference regions and segmentation strategies, with artificial intelligence, radiomics and automated segmentation potentially improving the integration of multimodal and quantitative assessment.
This review focuses on studies published over the past 18 months and examines the role of molecular PET imaging in differentiating treatment-related changes, especially pseudoprogression, from true tumor progression in gliomas, suggesting that static amino acid PET, particularly [18F]FET, may provide diagnostic information comparable to dynamic acquisition while simplifying protocols, that [68Ga]FAPI PET may add information on the tumor microenvironment, and that quantitative PET parameters are affected by reconstruction algorithms, reference regions and segmentation strategies, with artificial intelligence, radiomics and automated segmentation potentially improving the integration of multimodal and quantitative assessment.
This review focuses on studies published over the past 18 months and examines the role of molecular PET imaging in differentiating treatment-related changes, especially pseudoprogression, from true tumor progression in gliomas, suggesting that static amino acid PET, particularly [18F]FET, may provide diagnostic information comparable to dynamic acquisition while simplifying protocols, that [68Ga]FAPI PET may add information on the tumor microenvironment, and that quantitative PET parameters are affected by reconstruction algorithms, reference regions and segmentation strategies, with artificial intelligence, radiomics and automated segmentation potentially improving the integration of multimodal and quantitative assessment.
This review focuses on studies published over the past 18 months and examines the role of molecular PET imaging in differentiating treatment-related changes, especially pseudoprogression, from true tumor progression in gliomas, suggesting that static amino acid PET, particularly [18F]FET, may provide diagnostic information comparable to dynamic acquisition while simplifying protocols, that [68Ga]FAPI PET may add information on the tumor microenvironment, and that quantitative PET parameters are affected by reconstruction algorithms, reference regions and segmentation strategies, with artificial intelligence, radiomics and automated segmentation potentially improving the integration of multimodal and quantitative assessment.
FEBS letters This review surveys the landscape of deep learning-driven protein design pipelines, discusses tailored applications in peptide, small molecule, binder, vaccine, and antibody design, argues that current confidence metrics for filtering and evaluating designs remain optimized for static protein interfaces and can fail on underrepresented or conformationally complex targets, proposes ensemble-based methods as a promising avenue for improving design success rates, and highlights emerging strategies such as fold-switching scaffolds and molecular glues realized through engineered cyclic peptides that expand the functional scope of designed proteins.
This review surveys the landscape of deep learning-driven protein design pipelines, discusses tailored applications in peptide, small molecule, binder, vaccine, and antibody design, argues that current confidence metrics for filtering and evaluating designs remain optimized for static protein interfaces and can fail on underrepresented or conformationally complex targets, proposes ensemble-based methods as a promising avenue for improving design success rates, and highlights emerging strategies such as fold-switching scaffolds and molecular glues realized through engineered cyclic peptides that expand the functional scope of designed proteins.
This review surveys the landscape of deep learning-driven protein design pipelines, discusses tailored applications in peptide, small molecule, binder, vaccine, and antibody design, argues that current confidence metrics for filtering and evaluating designs remain optimized for static protein interfaces and can fail on underrepresented or conformationally complex targets, proposes ensemble-based methods as a promising avenue for improving design success rates, and highlights emerging strategies such as fold-switching scaffolds and molecular glues realized through engineered cyclic peptides that expand the functional scope of designed proteins.
This review surveys the landscape of deep learning-driven protein design pipelines, discusses tailored applications in peptide, small molecule, binder, vaccine, and antibody design, argues that current confidence metrics for filtering and evaluating designs remain optimized for static protein interfaces and can fail on underrepresented or conformationally complex targets, proposes ensemble-based methods as a promising avenue for improving design success rates, and highlights emerging strategies such as fold-switching scaffolds and molecular glues realized through engineered cyclic peptides that expand the functional scope of designed proteins.
Journal of the American College of Surgeons Using clinical notes from 105 patients in the NSQIP Breast Reconstruction pilot program (July 1, 2024–February 28, 2025), manually de-identified and processed with a customized ChatGPT 4.1 workflow targeting individual variables against a faculty plastic surgeon reference standard, this study evaluated 9,048 data points and found overall abstraction accuracy of 99.33% (61 errors) for the LLM versus 98.19% (164 errors) for human abstraction, with McNemar and Chi-square p<0.001; the LLM exceeded human abstraction for operative and postoperative variables but was slightly lower for preoperative variables, and the most frequent LLM errors involved prior breast surgical history (29/61) and prepectoral versus subpectoral implant or expander placement.
Using clinical notes from 105 patients in the NSQIP Breast Reconstruction pilot program (July 1, 2024–February 28, 2025), manually de-identified and processed with a customized ChatGPT 4.1 workflow targeting individual variables against a faculty plastic surgeon reference standard, this study evaluated 9,048 data points and found overall abstraction accuracy of 99.33% (61 errors) for the LLM versus 98.19% (164 errors) for human abstraction, with McNemar and Chi-square p<0.001; the LLM exceeded human abstraction for operative and postoperative variables but was slightly lower for preoperative variables, and the most frequent LLM errors involved prior breast surgical history (29/61) and prepectoral versus subpectoral implant or expander placement.
Using clinical notes from 105 patients in the NSQIP Breast Reconstruction pilot program (July 1, 2024–February 28, 2025), manually de-identified and processed with a customized ChatGPT 4.1 workflow targeting individual variables against a faculty plastic surgeon reference standard, this study evaluated 9,048 data points and found overall abstraction accuracy of 99.33% (61 errors) for the LLM versus 98.19% (164 errors) for human abstraction, with McNemar and Chi-square p<0.001; the LLM exceeded human abstraction for operative and postoperative variables but was slightly lower for preoperative variables, and the most frequent LLM errors involved prior breast surgical history (29/61) and prepectoral versus subpectoral implant or expander placement.
Using clinical notes from 105 patients in the NSQIP Breast Reconstruction pilot program (July 1, 2024–February 28, 2025), manually de-identified and processed with a customized ChatGPT 4.1 workflow targeting individual variables against a faculty plastic surgeon reference standard, this study evaluated 9,048 data points and found overall abstraction accuracy of 99.33% (61 errors) for the LLM versus 98.19% (164 errors) for human abstraction, with McNemar and Chi-square p<0.001; the LLM exceeded human abstraction for operative and postoperative variables but was slightly lower for preoperative variables, and the most frequent LLM errors involved prior breast surgical history (29/61) and prepectoral versus subpectoral implant or expander placement.
Pharmaceutical development and technology This review systematically surveys recent advances in natural biodegradable polymer-based microneedle drug delivery systems in terms of material selection, fabrication strategies, and controlled release mechanisms, and introduces a rational design framework that systematically integrates therapeutic objectives, polymer properties, mechanical performance, and release kinetics, while discussing translational challenges such as mechanical limitations, manufacturing scalability, stability concerns, and regulatory considerations, as well as emerging trends including stimuli-responsive systems, nanocarrier integration, personalized drug delivery, and data-driven strategies such as artificial intelligence and machine learning, arguing for a shift from empirical formulation toward a predictive, en
This review systematically surveys recent advances in natural biodegradable polymer-based microneedle drug delivery systems in terms of material selection, fabrication strategies, and controlled release mechanisms, and introduces a rational design framework that systematically integrates therapeutic objectives, polymer properties, mechanical performance, and release kinetics, while discussing translational challenges such as mechanical limitations, manufacturing scalability, stability concerns, and regulatory considerations, as well as emerging trends including stimuli-responsive systems, nanocarrier integration, personalized drug delivery, and data-driven strategies such as artificial intelligence and machine learning, arguing for a shift from empirical formulation toward a predictive, en
This review systematically surveys recent advances in natural biodegradable polymer-based microneedle drug delivery systems in terms of material selection, fabrication strategies, and controlled release mechanisms, and introduces a rational design framework that systematically integrates therapeutic objectives, polymer properties, mechanical performance, and release kinetics, while discussing translational challenges such as mechanical limitations, manufacturing scalability, stability concerns, and regulatory considerations, as well as emerging trends including stimuli-responsive systems, nanocarrier integration, personalized drug delivery, and data-driven strategies such as artificial intelligence and machine learning, arguing for a shift from empirical formulation toward a predictive, en
This review systematically surveys recent advances in natural biodegradable polymer-based microneedle drug delivery systems in terms of material selection, fabrication strategies, and controlled release mechanisms, and introduces a rational design framework that systematically integrates therapeutic objectives, polymer properties, mechanical performance, and release kinetics, while discussing translational challenges such as mechanical limitations, manufacturing scalability, stability concerns, and regulatory considerations, as well as emerging trends including stimuli-responsive systems, nanocarrier integration, personalized drug delivery, and data-driven strategies such as artificial intelligence and machine learning, arguing for a shift from empirical formulation toward a predictive, en
The Cleft palate-craniofacial journal : official publication of the American Cleft Palate-Craniofacial Association In this prospective, single-center study, speech samples from 40 children aged 2 to 17 (including individuals with velopharyngeal dysfunction, conditions associated with VPD, and healthy participants) were collected during speech-language pathologist-guided evaluation with consensus CAPS-A-AM ratings, and mel spectrograms of high vowels /i/ and /u/ from sustained vowels, isolated words, and sentences were used to train logistic regression, an EfficientNet-V2-S attention multiple-instance-learning CNN, and a CNN-XGBoost hybrid for binary hypernasality classification at two CAPS-A-AM thresholds (absent 0 versus any hypernasality 1-4, and absent/borderline 0-1 versus mild-to-severe 2-4); multiple independent modeling approaches detected clinically rated hypernasality, with EfficientNet-V2-S a
In this prospective, single-center study, speech samples from 40 children aged 2 to 17 (including individuals with velopharyngeal dysfunction, conditions associated with VPD, and healthy participants) were collected during speech-language pathologist-guided evaluation with consensus CAPS-A-AM ratings, and mel spectrograms of high vowels /i/ and /u/ from sustained vowels, isolated words, and sentences were used to train logistic regression, an EfficientNet-V2-S attention multiple-instance-learning CNN, and a CNN-XGBoost hybrid for binary hypernasality classification at two CAPS-A-AM thresholds (absent 0 versus any hypernasality 1-4, and absent/borderline 0-1 versus mild-to-severe 2-4); multiple independent modeling approaches detected clinically rated hypernasality, with EfficientNet-V2-S a
In this prospective, single-center study, speech samples from 40 children aged 2 to 17 (including individuals with velopharyngeal dysfunction, conditions associated with VPD, and healthy participants) were collected during speech-language pathologist-guided evaluation with consensus CAPS-A-AM ratings, and mel spectrograms of high vowels /i/ and /u/ from sustained vowels, isolated words, and sentences were used to train logistic regression, an EfficientNet-V2-S attention multiple-instance-learning CNN, and a CNN-XGBoost hybrid for binary hypernasality classification at two CAPS-A-AM thresholds (absent 0 versus any hypernasality 1-4, and absent/borderline 0-1 versus mild-to-severe 2-4); multiple independent modeling approaches detected clinically rated hypernasality, with EfficientNet-V2-S a
In this prospective, single-center study, speech samples from 40 children aged 2 to 17 (including individuals with velopharyngeal dysfunction, conditions associated with VPD, and healthy participants) were collected during speech-language pathologist-guided evaluation with consensus CAPS-A-AM ratings, and mel spectrograms of high vowels /i/ and /u/ from sustained vowels, isolated words, and sentences were used to train logistic regression, an EfficientNet-V2-S attention multiple-instance-learning CNN, and a CNN-XGBoost hybrid for binary hypernasality classification at two CAPS-A-AM thresholds (absent 0 versus any hypernasality 1-4, and absent/borderline 0-1 versus mild-to-severe 2-4); multiple independent modeling approaches detected clinically rated hypernasality, with EfficientNet-V2-S a
bioRxiv Using the multi-agent system ARES to analyze 1,552 TF binding datasets across 10 cell types and test competing mechanisms of TF-motif dependencies in specific cellular contexts against multi-omic data, the study found that inferred mechanisms converge on three operating routes—direct sequence recognition, protein-mediated recruitment or exclusion, and regulatory context—and that predictive motifs of target TF binding were read by their conventionally 'canonical' TFs in only one third of resolved dependencies, with motif similarity associated with shared regulatory region type but not shared transcriptional outcome, whereas reader identity was associated with both, thereby separating motif identity, reader identity, and regulatory output.
Using the multi-agent system ARES to analyze 1,552 TF binding datasets across 10 cell types and test competing mechanisms of TF-motif dependencies in specific cellular contexts against multi-omic data, the study found that inferred mechanisms converge on three operating routes—direct sequence recognition, protein-mediated recruitment or exclusion, and regulatory context—and that predictive motifs of target TF binding were read by their conventionally 'canonical' TFs in only one third of resolved dependencies, with motif similarity associated with shared regulatory region type but not shared transcriptional outcome, whereas reader identity was associated with both, thereby separating motif identity, reader identity, and regulatory output.
Using the multi-agent system ARES to analyze 1,552 TF binding datasets across 10 cell types and test competing mechanisms of TF-motif dependencies in specific cellular contexts against multi-omic data, the study found that inferred mechanisms converge on three operating routes—direct sequence recognition, protein-mediated recruitment or exclusion, and regulatory context—and that predictive motifs of target TF binding were read by their conventionally 'canonical' TFs in only one third of resolved dependencies, with motif similarity associated with shared regulatory region type but not shared transcriptional outcome, whereas reader identity was associated with both, thereby separating motif identity, reader identity, and regulatory output.
Using the multi-agent system ARES to analyze 1,552 TF binding datasets across 10 cell types and test competing mechanisms of TF-motif dependencies in specific cellular contexts against multi-omic data, the study found that inferred mechanisms converge on three operating routes—direct sequence recognition, protein-mediated recruitment or exclusion, and regulatory context—and that predictive motifs of target TF binding were read by their conventionally 'canonical' TFs in only one third of resolved dependencies, with motif similarity associated with shared regulatory region type but not shared transcriptional outcome, whereas reader identity was associated with both, thereby separating motif identity, reader identity, and regulatory output.
International journal of ophthalmology This systematic review evaluates 34 studies from 2018 to 2025, finding that AI often exceeds 90% accuracy and can match or outperform expert clinicians in diagnosing common ocular diseases such as diabetic retinopathy, glaucoma, retinopathy of prematurity, and age-related macular degeneration, yet real-world deployment remains constrained by three gaps—disjointed integration into clinical workflows, lack of transparency in AI decision-making, and poor generalizability across diverse populations—and it proposes actionable pathways to bridge the "last-mile gap" between research and clinical practice.
This systematic review evaluates 34 studies from 2018 to 2025, finding that AI often exceeds 90% accuracy and can match or outperform expert clinicians in diagnosing common ocular diseases such as diabetic retinopathy, glaucoma, retinopathy of prematurity, and age-related macular degeneration, yet real-world deployment remains constrained by three gaps—disjointed integration into clinical workflows, lack of transparency in AI decision-making, and poor generalizability across diverse populations—and it proposes actionable pathways to bridge the "last-mile gap" between research and clinical practice.
This systematic review evaluates 34 studies from 2018 to 2025, finding that AI often exceeds 90% accuracy and can match or outperform expert clinicians in diagnosing common ocular diseases such as diabetic retinopathy, glaucoma, retinopathy of prematurity, and age-related macular degeneration, yet real-world deployment remains constrained by three gaps—disjointed integration into clinical workflows, lack of transparency in AI decision-making, and poor generalizability across diverse populations—and it proposes actionable pathways to bridge the "last-mile gap" between research and clinical practice.
This systematic review evaluates 34 studies from 2018 to 2025, finding that AI often exceeds 90% accuracy and can match or outperform expert clinicians in diagnosing common ocular diseases such as diabetic retinopathy, glaucoma, retinopathy of prematurity, and age-related macular degeneration, yet real-world deployment remains constrained by three gaps—disjointed integration into clinical workflows, lack of transparency in AI decision-making, and poor generalizability across diverse populations—and it proposes actionable pathways to bridge the "last-mile gap" between research and clinical practice.
Journal of gastroenterology and hepatology This narrative review is the first to synthesize the convergence of artificial intelligence and endohepatology into four functional pillars—intelligent hemodynamic assessment, virtual histology, precision tissue acquisition, and integrated risk stratification with therapeutic and decision support—and proposes a three-tier readiness framework to separate demonstrated capabilities from extrapolated and conceptual applications, concluding that the field is still very early-stage but that AI has strong potential to turn endohepatology into a single machine-driven diagnostic and therapeutic platform provided standardized datasets, prospective validation, and clear regulatory and governance standards are in place.
This narrative review is the first to synthesize the convergence of artificial intelligence and endohepatology into four functional pillars—intelligent hemodynamic assessment, virtual histology, precision tissue acquisition, and integrated risk stratification with therapeutic and decision support—and proposes a three-tier readiness framework to separate demonstrated capabilities from extrapolated and conceptual applications, concluding that the field is still very early-stage but that AI has strong potential to turn endohepatology into a single machine-driven diagnostic and therapeutic platform provided standardized datasets, prospective validation, and clear regulatory and governance standards are in place.
This narrative review is the first to synthesize the convergence of artificial intelligence and endohepatology into four functional pillars—intelligent hemodynamic assessment, virtual histology, precision tissue acquisition, and integrated risk stratification with therapeutic and decision support—and proposes a three-tier readiness framework to separate demonstrated capabilities from extrapolated and conceptual applications, concluding that the field is still very early-stage but that AI has strong potential to turn endohepatology into a single machine-driven diagnostic and therapeutic platform provided standardized datasets, prospective validation, and clear regulatory and governance standards are in place.
This narrative review is the first to synthesize the convergence of artificial intelligence and endohepatology into four functional pillars—intelligent hemodynamic assessment, virtual histology, precision tissue acquisition, and integrated risk stratification with therapeutic and decision support—and proposes a three-tier readiness framework to separate demonstrated capabilities from extrapolated and conceptual applications, concluding that the field is still very early-stage but that AI has strong potential to turn endohepatology into a single machine-driven diagnostic and therapeutic platform provided standardized datasets, prospective validation, and clear regulatory and governance standards are in place.
Bioinformatics (Oxford, England) The work presents scHPGT, a single-cell Heterogeneous Prior-Guided Transformer that integrates unpaired RNA and chromatin accessibility profiles in a shared latent space using modality-specific encoders, a prior-guided cross-modal Transformer that constrains gene-peak attention with regulatory links, and a domain-adversarial objective, improving clustering agreement, label transfer and biological structure preservation across PBMC3k, mouse spleen, CITE-seq/ASAP-seq PBMC and PBMC10k benchmarks while avoiding forced correspondence of unmatched or condition-specific states in partial-overlap and condition-shift settings, with attention-derived links recovering regulatory relationships, marker-gene regulatory regions, transcription factor programs and regulatory activity profiles.
The work presents scHPGT, a single-cell Heterogeneous Prior-Guided Transformer that integrates unpaired RNA and chromatin accessibility profiles in a shared latent space using modality-specific encoders, a prior-guided cross-modal Transformer that constrains gene-peak attention with regulatory links, and a domain-adversarial objective, improving clustering agreement, label transfer and biological structure preservation across PBMC3k, mouse spleen, CITE-seq/ASAP-seq PBMC and PBMC10k benchmarks while avoiding forced correspondence of unmatched or condition-specific states in partial-overlap and condition-shift settings, with attention-derived links recovering regulatory relationships, marker-gene regulatory regions, transcription factor programs and regulatory activity profiles.
The work presents scHPGT, a single-cell Heterogeneous Prior-Guided Transformer that integrates unpaired RNA and chromatin accessibility profiles in a shared latent space using modality-specific encoders, a prior-guided cross-modal Transformer that constrains gene-peak attention with regulatory links, and a domain-adversarial objective, improving clustering agreement, label transfer and biological structure preservation across PBMC3k, mouse spleen, CITE-seq/ASAP-seq PBMC and PBMC10k benchmarks while avoiding forced correspondence of unmatched or condition-specific states in partial-overlap and condition-shift settings, with attention-derived links recovering regulatory relationships, marker-gene regulatory regions, transcription factor programs and regulatory activity profiles.
The work presents scHPGT, a single-cell Heterogeneous Prior-Guided Transformer that integrates unpaired RNA and chromatin accessibility profiles in a shared latent space using modality-specific encoders, a prior-guided cross-modal Transformer that constrains gene-peak attention with regulatory links, and a domain-adversarial objective, improving clustering agreement, label transfer and biological structure preservation across PBMC3k, mouse spleen, CITE-seq/ASAP-seq PBMC and PBMC10k benchmarks while avoiding forced correspondence of unmatched or condition-specific states in partial-overlap and condition-shift settings, with attention-derived links recovering regulatory relationships, marker-gene regulatory regions, transcription factor programs and regulatory activity profiles.
bioRxiv Using the tabular foundation model TabPFN informed by a small library of living hydrogels, this work predicts macroscopic material properties of Escherichia coli-produced living hydrogels containing CsgA-based fibres fused to genetically encoded PEG-like biopolymers from genetic and process parameters, achieving the strongest prediction for storage modulus G' (R2 = 85.1%) on an independent validation set with a 48.0% RMSE reduction versus linear regression, and further enabling property-guided design to identify parameters for desired properties.
Using the tabular foundation model TabPFN informed by a small library of living hydrogels, this work predicts macroscopic material properties of Escherichia coli-produced living hydrogels containing CsgA-based fibres fused to genetically encoded PEG-like biopolymers from genetic and process parameters, achieving the strongest prediction for storage modulus G' (R2 = 85.1%) on an independent validation set with a 48.0% RMSE reduction versus linear regression, and further enabling property-guided design to identify parameters for desired properties.
Using the tabular foundation model TabPFN informed by a small library of living hydrogels, this work predicts macroscopic material properties of Escherichia coli-produced living hydrogels containing CsgA-based fibres fused to genetically encoded PEG-like biopolymers from genetic and process parameters, achieving the strongest prediction for storage modulus G' (R2 = 85.1%) on an independent validation set with a 48.0% RMSE reduction versus linear regression, and further enabling property-guided design to identify parameters for desired properties.
Using the tabular foundation model TabPFN informed by a small library of living hydrogels, this work predicts macroscopic material properties of Escherichia coli-produced living hydrogels containing CsgA-based fibres fused to genetically encoded PEG-like biopolymers from genetic and process parameters, achieving the strongest prediction for storage modulus G' (R2 = 85.1%) on an independent validation set with a 48.0% RMSE reduction versus linear regression, and further enabling property-guided design to identify parameters for desired properties.
bioRxiv This work builds BoneGraph, a bone-science-specific system delivered as a five-tab web application over a shared substrate of 7,449 documents, 248,629 SPECTER2 passage vectors, and a bone knowledge graph of 1,597 concepts and 1,699 causal relations, reporting MRR 0.928 on a 30-question seven-domain retrieval benchmark, 92.6% accuracy for the bone-region classifier on held-out MURA, and an increase in answer accuracy from 42% to 78% when the correct passage is supplied, with all inference performed locally on a single NVIDIA Jetson AGX Orin and no third-party API calls.
This work builds BoneGraph, a bone-science-specific system delivered as a five-tab web application over a shared substrate of 7,449 documents, 248,629 SPECTER2 passage vectors, and a bone knowledge graph of 1,597 concepts and 1,699 causal relations, reporting MRR 0.928 on a 30-question seven-domain retrieval benchmark, 92.6% accuracy for the bone-region classifier on held-out MURA, and an increase in answer accuracy from 42% to 78% when the correct passage is supplied, with all inference performed locally on a single NVIDIA Jetson AGX Orin and no third-party API calls.
This work builds BoneGraph, a bone-science-specific system delivered as a five-tab web application over a shared substrate of 7,449 documents, 248,629 SPECTER2 passage vectors, and a bone knowledge graph of 1,597 concepts and 1,699 causal relations, reporting MRR 0.928 on a 30-question seven-domain retrieval benchmark, 92.6% accuracy for the bone-region classifier on held-out MURA, and an increase in answer accuracy from 42% to 78% when the correct passage is supplied, with all inference performed locally on a single NVIDIA Jetson AGX Orin and no third-party API calls.
This work builds BoneGraph, a bone-science-specific system delivered as a five-tab web application over a shared substrate of 7,449 documents, 248,629 SPECTER2 passage vectors, and a bone knowledge graph of 1,597 concepts and 1,699 causal relations, reporting MRR 0.928 on a 30-question seven-domain retrieval benchmark, 92.6% accuracy for the bone-region classifier on held-out MURA, and an increase in answer accuracy from 42% to 78% when the correct passage is supplied, with all inference performed locally on a single NVIDIA Jetson AGX Orin and no third-party API calls.
Medical Science Monitor This narrative review synthesizes recent clinical studies and technological advances to map virtual reality (VR) applications in otolaryngology nursing across patient-facing uses (preoperative education, pain and anxiety relief, vestibular rehabilitation, swallowing assessment and therapy, hearing rehabilitation, and tinnitus management) and nurse-facing uses (professional training and skills development), reporting that vestibular rehabilitation is supported by strong evidence from randomized controlled trials while swallowing therapy and tinnitus management rest mainly on preliminary exploratory studies, and discussing technical, economic, and clinical barriers alongside prospects for integrating VR with artificial intelligence, wearable biosensors, and telerehabilitation platforms.
This narrative review synthesizes recent clinical studies and technological advances to map virtual reality (VR) applications in otolaryngology nursing across patient-facing uses (preoperative education, pain and anxiety relief, vestibular rehabilitation, swallowing assessment and therapy, hearing rehabilitation, and tinnitus management) and nurse-facing uses (professional training and skills development), reporting that vestibular rehabilitation is supported by strong evidence from randomized controlled trials while swallowing therapy and tinnitus management rest mainly on preliminary exploratory studies, and discussing technical, economic, and clinical barriers alongside prospects for integrating VR with artificial intelligence, wearable biosensors, and telerehabilitation platforms.
This narrative review synthesizes recent clinical studies and technological advances to map virtual reality (VR) applications in otolaryngology nursing across patient-facing uses (preoperative education, pain and anxiety relief, vestibular rehabilitation, swallowing assessment and therapy, hearing rehabilitation, and tinnitus management) and nurse-facing uses (professional training and skills development), reporting that vestibular rehabilitation is supported by strong evidence from randomized controlled trials while swallowing therapy and tinnitus management rest mainly on preliminary exploratory studies, and discussing technical, economic, and clinical barriers alongside prospects for integrating VR with artificial intelligence, wearable biosensors, and telerehabilitation platforms.
This narrative review synthesizes recent clinical studies and technological advances to map virtual reality (VR) applications in otolaryngology nursing across patient-facing uses (preoperative education, pain and anxiety relief, vestibular rehabilitation, swallowing assessment and therapy, hearing rehabilitation, and tinnitus management) and nurse-facing uses (professional training and skills development), reporting that vestibular rehabilitation is supported by strong evidence from randomized controlled trials while swallowing therapy and tinnitus management rest mainly on preliminary exploratory studies, and discussing technical, economic, and clinical barriers alongside prospects for integrating VR with artificial intelligence, wearable biosensors, and telerehabilitation platforms.
bioRxiv This work introduces TRACEDD, a tool-first multi-agent framework in which large language models serve as the reasoning and orchestration layer while domain-validated computational tools supply structural, chemical, pharmacological, and synthetic predictions, and it demonstrates an end-to-end JAK2 workflow spanning target validation, structure retrieval or AlphaFold prediction, druggable pocket identification, reinforcement-learning-based de novo molecular generation, lead optimization, ADMET and bioactivity evaluation, literature evidence integration, and retrosynthesis planning, with each decision linked to explicit tool invocations and intermediate evidence.
This work introduces TRACEDD, a tool-first multi-agent framework in which large language models serve as the reasoning and orchestration layer while domain-validated computational tools supply structural, chemical, pharmacological, and synthetic predictions, and it demonstrates an end-to-end JAK2 workflow spanning target validation, structure retrieval or AlphaFold prediction, druggable pocket identification, reinforcement-learning-based de novo molecular generation, lead optimization, ADMET and bioactivity evaluation, literature evidence integration, and retrosynthesis planning, with each decision linked to explicit tool invocations and intermediate evidence.
This work introduces TRACEDD, a tool-first multi-agent framework in which large language models serve as the reasoning and orchestration layer while domain-validated computational tools supply structural, chemical, pharmacological, and synthetic predictions, and it demonstrates an end-to-end JAK2 workflow spanning target validation, structure retrieval or AlphaFold prediction, druggable pocket identification, reinforcement-learning-based de novo molecular generation, lead optimization, ADMET and bioactivity evaluation, literature evidence integration, and retrosynthesis planning, with each decision linked to explicit tool invocations and intermediate evidence.
This work introduces TRACEDD, a tool-first multi-agent framework in which large language models serve as the reasoning and orchestration layer while domain-validated computational tools supply structural, chemical, pharmacological, and synthetic predictions, and it demonstrates an end-to-end JAK2 workflow spanning target validation, structure retrieval or AlphaFold prediction, druggable pocket identification, reinforcement-learning-based de novo molecular generation, lead optimization, ADMET and bioactivity evaluation, literature evidence integration, and retrosynthesis planning, with each decision linked to explicit tool invocations and intermediate evidence.
Google Research This research experiment lets teachers enter a curriculum topic and have generative UI, under instructional-design guardrails, dynamically produce interactive simulations with leveled challenges, hints, feedback, and worked solutions, and it releases a library of over 30 teacher-reviewed STEM learning interactives; UK STEM teachers rated 40 interactives overall as good or excellent, with physics and chemistry most amenable to simulation creation, and 12 US teachers who each requested three custom interactives gave an average quality rating of 8 out of 10.
This research experiment lets teachers enter a curriculum topic and have generative UI, under instructional-design guardrails, dynamically produce interactive simulations with leveled challenges, hints, feedback, and worked solutions, and it releases a library of over 30 teacher-reviewed STEM learning interactives; UK STEM teachers rated 40 interactives overall as good or excellent, with physics and chemistry most amenable to simulation creation, and 12 US teachers who each requested three custom interactives gave an average quality rating of 8 out of 10.
This research experiment lets teachers enter a curriculum topic and have generative UI, under instructional-design guardrails, dynamically produce interactive simulations with leveled challenges, hints, feedback, and worked solutions, and it releases a library of over 30 teacher-reviewed STEM learning interactives; UK STEM teachers rated 40 interactives overall as good or excellent, with physics and chemistry most amenable to simulation creation, and 12 US teachers who each requested three custom interactives gave an average quality rating of 8 out of 10.
This research experiment lets teachers enter a curriculum topic and have generative UI, under instructional-design guardrails, dynamically produce interactive simulations with leveled challenges, hints, feedback, and worked solutions, and it releases a library of over 30 teacher-reviewed STEM learning interactives; UK STEM teachers rated 40 interactives overall as good or excellent, with physics and chemistry most amenable to simulation creation, and 12 US teachers who each requested three custom interactives gave an average quality rating of 8 out of 10.
Terence Tao blog RSS This opinion piece by Dimitris Koukoulopoulos, published on Terence Tao's blog, takes OpenAI's announced solution to the Navier-Stokes Millennium Prize problem and the controversy over how it was obtained and released as a wake-up call, and argues that the scientific community should quickly build publicly funded frontier AI systems and compute infrastructure, a "CERN for AI-assisted science," offering every researcher free baseline access to conversational and agentic interfaces while allocating large-scale compute and multi-agent systems through competitive applications, so that research tools do not remain concentrated in a handful of private companies and a two-tier scientific system does not emerge.
This opinion piece by Dimitris Koukoulopoulos, published on Terence Tao's blog, takes OpenAI's announced solution to the Navier-Stokes Millennium Prize problem and the controversy over how it was obtained and released as a wake-up call, and argues that the scientific community should quickly build publicly funded frontier AI systems and compute infrastructure, a "CERN for AI-assisted science," offering every researcher free baseline access to conversational and agentic interfaces while allocating large-scale compute and multi-agent systems through competitive applications, so that research tools do not remain concentrated in a handful of private companies and a two-tier scientific system does not emerge.
This opinion piece by Dimitris Koukoulopoulos, published on Terence Tao's blog, takes OpenAI's announced solution to the Navier-Stokes Millennium Prize problem and the controversy over how it was obtained and released as a wake-up call, and argues that the scientific community should quickly build publicly funded frontier AI systems and compute infrastructure, a "CERN for AI-assisted science," offering every researcher free baseline access to conversational and agentic interfaces while allocating large-scale compute and multi-agent systems through competitive applications, so that research tools do not remain concentrated in a handful of private companies and a two-tier scientific system does not emerge.
This opinion piece by Dimitris Koukoulopoulos, published on Terence Tao's blog, takes OpenAI's announced solution to the Navier-Stokes Millennium Prize problem and the controversy over how it was obtained and released as a wake-up call, and argues that the scientific community should quickly build publicly funded frontier AI systems and compute infrastructure, a "CERN for AI-assisted science," offering every researcher free baseline access to conversational and agentic interfaces while allocating large-scale compute and multi-agent systems through competitive applications, so that research tools do not remain concentrated in a handful of private companies and a two-tier scientific system does not emerge.
Nature News Using tissue staining and RNA sequencing of mouse embryos 7.5 days after conception, red fluorescent lineage tracing, directed differentiation of human pluripotent stem cells, and a search across monkeys, chickens, zebrafish and even acorn worms, this work proposes that the brain is not made by a single type of starter cell but by two non-mixing progenitor populations, one forming the hindbrain and the other the forebrain and midbrain, and it establishes an efficient way to coax stem cells into hindbrain motor neurons.
Using tissue staining and RNA sequencing of mouse embryos 7.5 days after conception, red fluorescent lineage tracing, directed differentiation of human pluripotent stem cells, and a search across monkeys, chickens, zebrafish and even acorn worms, this work proposes that the brain is not made by a single type of starter cell but by two non-mixing progenitor populations, one forming the hindbrain and the other the forebrain and midbrain, and it establishes an efficient way to coax stem cells into hindbrain motor neurons.
Using tissue staining and RNA sequencing of mouse embryos 7.5 days after conception, red fluorescent lineage tracing, directed differentiation of human pluripotent stem cells, and a search across monkeys, chickens, zebrafish and even acorn worms, this work proposes that the brain is not made by a single type of starter cell but by two non-mixing progenitor populations, one forming the hindbrain and the other the forebrain and midbrain, and it establishes an efficient way to coax stem cells into hindbrain motor neurons.
Using tissue staining and RNA sequencing of mouse embryos 7.5 days after conception, red fluorescent lineage tracing, directed differentiation of human pluripotent stem cells, and a search across monkeys, chickens, zebrafish and even acorn worms, this work proposes that the brain is not made by a single type of starter cell but by two non-mixing progenitor populations, one forming the hindbrain and the other the forebrain and midbrain, and it establishes an efficient way to coax stem cells into hindbrain motor neurons.
Nature News According to a Nature news report, OpenAI claims its most advanced AI model proved that the Navier–Stokes equations produce a 'singularity' in some instances, predicting physically impossible infinite speeds; the report situates that claim by reviewing the equations' known limits for compressible fluids and rarefied gases, and alternative modelling routes such as the Boltzmann equation, molecular simulation, and a 'triple decker' multiscale coupling.
According to a Nature news report, OpenAI claims its most advanced AI model proved that the Navier–Stokes equations produce a 'singularity' in some instances, predicting physically impossible infinite speeds; the report situates that claim by reviewing the equations' known limits for compressible fluids and rarefied gases, and alternative modelling routes such as the Boltzmann equation, molecular simulation, and a 'triple decker' multiscale coupling.
According to a Nature news report, OpenAI claims its most advanced AI model proved that the Navier–Stokes equations produce a 'singularity' in some instances, predicting physically impossible infinite speeds; the report situates that claim by reviewing the equations' known limits for compressible fluids and rarefied gases, and alternative modelling routes such as the Boltzmann equation, molecular simulation, and a 'triple decker' multiscale coupling.
According to a Nature news report, OpenAI claims its most advanced AI model proved that the Navier–Stokes equations produce a 'singularity' in some instances, predicting physically impossible infinite speeds; the report situates that claim by reviewing the equations' known limits for compressible fluids and rarefied gases, and alternative modelling routes such as the Boltzmann equation, molecular simulation, and a 'triple decker' multiscale coupling.
Nature News This report surveys Europe's latest moves toward space sovereignty after the inaugural International Space Summit in Paris, including German company Isar Aerospace's 5 September launch from Andøya in Norway described as the first commercial rocket to reach orbit from continental Europe, the EU's planned 348-satellite IRIS 2 constellation, Germany's plan to invest about €35 billion in military space equipment by 2030, and The Exploration Company's $450 million raise to build the Nyx spacecraft, while quoting European Commission president Ursula von der Leyen, French president Emmanuel Macron, former Eutelsat chief executive Eva Berneke and Durham University's Bleddyn Bowen on Europe's effort to reduce reliance on US launch and satellite-internet capabilities.
This report surveys Europe's latest moves toward space sovereignty after the inaugural International Space Summit in Paris, including German company Isar Aerospace's 5 September launch from Andøya in Norway described as the first commercial rocket to reach orbit from continental Europe, the EU's planned 348-satellite IRIS 2 constellation, Germany's plan to invest about €35 billion in military space equipment by 2030, and The Exploration Company's $450 million raise to build the Nyx spacecraft, while quoting European Commission president Ursula von der Leyen, French president Emmanuel Macron, former Eutelsat chief executive Eva Berneke and Durham University's Bleddyn Bowen on Europe's effort to reduce reliance on US launch and satellite-internet capabilities.
This report surveys Europe's latest moves toward space sovereignty after the inaugural International Space Summit in Paris, including German company Isar Aerospace's 5 September launch from Andøya in Norway described as the first commercial rocket to reach orbit from continental Europe, the EU's planned 348-satellite IRIS 2 constellation, Germany's plan to invest about €35 billion in military space equipment by 2030, and The Exploration Company's $450 million raise to build the Nyx spacecraft, while quoting European Commission president Ursula von der Leyen, French president Emmanuel Macron, former Eutelsat chief executive Eva Berneke and Durham University's Bleddyn Bowen on Europe's effort to reduce reliance on US launch and satellite-internet capabilities.
This report surveys Europe's latest moves toward space sovereignty after the inaugural International Space Summit in Paris, including German company Isar Aerospace's 5 September launch from Andøya in Norway described as the first commercial rocket to reach orbit from continental Europe, the EU's planned 348-satellite IRIS 2 constellation, Germany's plan to invest about €35 billion in military space equipment by 2030, and The Exploration Company's $450 million raise to build the Nyx spacecraft, while quoting European Commission president Ursula von der Leyen, French president Emmanuel Macron, former Eutelsat chief executive Eva Berneke and Durham University's Bleddyn Bowen on Europe's effort to reduce reliance on US launch and satellite-internet capabilities.
Nature News An analysis published in Science used 2010–2022 data from the China National Intellectual Property Administration, Web of Science, and the China Stock Market and Accounting Research Database to compare Chinese firms placed on the US Entity List with similar unsanctioned firms, finding that sanctioned firms produced 72.3% more patents citing at least one scientific publication and published 33.3% more papers indexed in the China National Knowledge Infrastructure and 85.2% more in Web of Science, while Chinese patents citing scientific literature overall rose from 5,225 in 2010 to 94,441 in 2022.
An analysis published in Science used 2010–2022 data from the China National Intellectual Property Administration, Web of Science, and the China Stock Market and Accounting Research Database to compare Chinese firms placed on the US Entity List with similar unsanctioned firms, finding that sanctioned firms produced 72.3% more patents citing at least one scientific publication and published 33.3% more papers indexed in the China National Knowledge Infrastructure and 85.2% more in Web of Science, while Chinese patents citing scientific literature overall rose from 5,225 in 2010 to 94,441 in 2022.
An analysis published in Science used 2010–2022 data from the China National Intellectual Property Administration, Web of Science, and the China Stock Market and Accounting Research Database to compare Chinese firms placed on the US Entity List with similar unsanctioned firms, finding that sanctioned firms produced 72.3% more patents citing at least one scientific publication and published 33.3% more papers indexed in the China National Knowledge Infrastructure and 85.2% more in Web of Science, while Chinese patents citing scientific literature overall rose from 5,225 in 2010 to 94,441 in 2022.
An analysis published in Science used 2010–2022 data from the China National Intellectual Property Administration, Web of Science, and the China Stock Market and Accounting Research Database to compare Chinese firms placed on the US Entity List with similar unsanctioned firms, finding that sanctioned firms produced 72.3% more patents citing at least one scientific publication and published 33.3% more papers indexed in the China National Knowledge Infrastructure and 85.2% more in Web of Science, while Chinese patents citing scientific literature overall rose from 5,225 in 2010 to 94,441 in 2022.
Nature News A man with a slowly progressing form of motor neuron disease (ALS) caused by a rare CHCHD10 mutation became the first person to receive an RNA antisense oligonucleotide therapy targeting his specific disease-causing mutation; after three 50-milligram and three 75-milligram doses delivered into his spine between April 2024 and April 2025, he had no serious side effects, and one year later his blood neurofilament light chain levels had fallen to the normal reference range, his motor, breathing and neurological function scores had improved, breathing and cognition scores remained stable, and he continued to work as a physician.
A man with a slowly progressing form of motor neuron disease (ALS) caused by a rare CHCHD10 mutation became the first person to receive an RNA antisense oligonucleotide therapy targeting his specific disease-causing mutation; after three 50-milligram and three 75-milligram doses delivered into his spine between April 2024 and April 2025, he had no serious side effects, and one year later his blood neurofilament light chain levels had fallen to the normal reference range, his motor, breathing and neurological function scores had improved, breathing and cognition scores remained stable, and he continued to work as a physician.
A man with a slowly progressing form of motor neuron disease (ALS) caused by a rare CHCHD10 mutation became the first person to receive an RNA antisense oligonucleotide therapy targeting his specific disease-causing mutation; after three 50-milligram and three 75-milligram doses delivered into his spine between April 2024 and April 2025, he had no serious side effects, and one year later his blood neurofilament light chain levels had fallen to the normal reference range, his motor, breathing and neurological function scores had improved, breathing and cognition scores remained stable, and he continued to work as a physician.
A man with a slowly progressing form of motor neuron disease (ALS) caused by a rare CHCHD10 mutation became the first person to receive an RNA antisense oligonucleotide therapy targeting his specific disease-causing mutation; after three 50-milligram and three 75-milligram doses delivered into his spine between April 2024 and April 2025, he had no serious side effects, and one year later his blood neurofilament light chain levels had fallen to the normal reference range, his motor, breathing and neurological function scores had improved, breathing and cognition scores remained stable, and he continued to work as a physician.