Sunyong Yoo
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Sunyong Yoo has a total of 114 co-authors. Here are the top co-authors by number of shared publications:
- Doheon Lee
Department of Bio and Brain Engineering, KAIST, Daejeon, Republic of Korea
11 shared publications Doheon Lee
Bio-Synergy Research Center, 34141 Daejeon, Republic of Korea
6 shared publicationsMoonshik Shin
Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology, Daejeon, South Korea
6 shared publications- Jinmyung Jung
Division of Data Science, College of Information and Communication Technology, The University of Suwon, Hwaseong 18323, Republic of Korea
5 shared publications - Gwan-Su Yi
Department of Bio and Brain Engineering, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong-gu, Daejeon 34141, Korea
4 shared publications - Hojung Nam
AI Graduate School, Gwangju Institute of Science and Technology (GIST), Gwangju, Republic of Korea
4 shared publications
Latest Publications
- Open Access
- 19 November 2026
- Published by Wolters Kluwer Health in Pediatric Critical Care Medicine
Long-Term Neurologic Complication and Healthcare Burden in Infants Receiving Extracorporeal Life Support: A Nationwide Population-Based Cohort Study
Objectives: To evaluate survival, neurologic complications, and healthcare burden in infant extracorporeal life support (ECLS) recipients using nationally representative administrative data from a single-payer healthcare system. Design: Retrospective, population-based cohort study. Setting: South Korea’s National Health Insurance Service, covering 97% of the population. Patients: Infants under 1 year who received ECLS between 2009 and 2022. Interventions: None. Measurements and Main Results: Of 698 infants who underwent ECLS, 212 infants (30.4%) survived during follow-up (estimated 1-yr mortality: 67.1%). Among survivors, 128 (60.4%) developed neurologic injuries, including acute events such as ischemic stroke, hemorrhagic stroke, and hypoxic-ischemic brain injury (HIBI), as well as long-term sequelae such as cerebral palsy, epilepsy, hearing loss, and developmental delay. Most acute neurologic injuries were diagnosed within the first year, whereas developmental delay showed prolonged diagnostic windows extending up to 12 years. Survivors with neurologic injury had significantly longer hospital stays (432.1 vs. 182.3 d; p < 0.001) and higher cumulative healthcare costs during long-term follow-up. Median index hospitalization costs were USD 36,000–37,000. The burden analysis showed distinct profiles across neurologic complications: developmental delay was the most prevalent sequela, whereas HIBI was associated with the highest cumulative healthcare cost despite lower prevalence. Conclusions: Neurologic complications following ECLS in infants are common and impose substantial long-term healthcare burden. These findings highlight the need for early risk stratification, structured neurologic and neurodevelopmental monitoring, and sustained rehabilitation planning in this high-risk population. This study provides a foundation for international benchmarking and health policy planning in pediatric critical care.
- 19 November 2026
- Published by Wolters Kluwer Health in Pediatric Critical Care Medicine
Open AccessLong-Term Neurologic Complication and Healthcare Burden in Infants Receiving Extracorporeal Life Support: A Nationwide Population-Based Cohort Study
Objectives: To evaluate survival, neurologic complications, and healthcare burden in infant extracorporeal life support (ECLS) recipients using nationally representative administrative data from a single-payer healthcare system. Design: Retrospective, population-based cohort study. Setting: South Korea’s National Health Insurance Service, covering 97% of the population. Patients: Infants under 1 year who received ECLS between 2009 and 2022. Interventions: None. Measurements and Main Results: Of 698 infants who underwent ECLS, 212 infants (30.4%) survived during follow-up (estimated 1-yr mortality: 67.1%). Among survivors, 128 (60.4%) developed neurologic injuries, including acute events such as ischemic stroke, hemorrhagic stroke, and hypoxic-ischemic brain injury (HIBI), as well as long-term sequelae such as cerebral palsy, epilepsy, hearing loss, and developmental delay. Most acute neurologic injuries were diagnosed within the first year, whereas developmental delay showed prolonged diagnostic windows extending up to 12 years. Survivors with neurologic injury had significantly longer hospital stays (432.1 vs. 182.3 d; p < 0.001) and higher cumulative healthcare costs during long-term follow-up. Median index hospitalization costs were USD 36,000–37,000. The burden analysis showed distinct profiles across neurologic complications: developmental delay was the most prevalent sequela, whereas HIBI was associated with the highest cumulative healthcare cost despite lower prevalence. Conclusions: Neurologic complications following ECLS in infants are common and impose substantial long-term healthcare burden. These findings highlight the need for early risk stratification, structured neurologic and neurodevelopmental monitoring, and sustained rehabilitation planning in this high-risk population. This study provides a foundation for international benchmarking and health policy planning in pediatric critical care.
- Open Access
- 31 August 2026
- Published by Frontiers Media SA in Frontiers in Pediatrics
- Vol. 14, 1934650
Virtual reality for rehabilitation and engagement in pediatric intensive care: a systematic review of feasibility, safety and outcomes
Children admitted to pediatric intensive care units (PICUs) are at risk of acquired weakness, delirium, and long-term functional and neurocognitive morbidity. Although early rehabilitation is recommended to mitigate these adverse outcomes, its delivery remains challenging. Virtual reality (VR) has been proposed as a means of enhancing engagement in rehabilitation in this setting. This systematic review evaluated the feasibility, safety, and outcomes of VR delivered to children during routine pediatric intensive care to support rehabilitation, participation in rehabilitation, or emotional regulation. Three databases were searched through 8 August 2026, and the reference list of one relevant scoping review was screened; two reviewers independently screened all records (Cohen's κ = 0.85). Six studies involving 188 children were included: one randomized pilot trial, one prospective single-arm trial, two cross-sectional single-arm studies, and two case reports. Risk of bias was high, serious, or critical in every appraised study, and the certainty of evidence was very low across all outcome domains. The randomized trial found higher consciousness scores at 2 weeks in the VR group but no difference in social and personal functional ability at 3 months. In uncontrolled studies, most children were rated as highly engaged and reported enjoying VR, while parents reported a calming effect more frequently than the children themselves; a transient increase in one heart rate variability parameter occurred during VR but was not sustained afterward. Adverse events were reported in four of the six studies at rates of 3.1% to 20%, comprising dizziness, nausea, visual discomfort, and mild irritability, with one serious event adjudicated as unrelated to VR; the two studies that applied pre-specified criteria recorded no events and one unrelated event, respectively. VR appears feasible and safe for hemodynamically stable, alert children in the PICU and is consistently associated with high engagement. However, whether VR improves functional outcomes remains uncertain. Eligibility, rather than tolerability, may be the principal barrier to implementation, and the available evidence applies to only a minority of the PICU population. Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/view/CRD42024607286 , PROSPERO, CRD42024607286
- 31 August 2026
- Published by Frontiers Media SA in Frontiers in Pediatrics
- Vol. 14, 1934650
Open AccessVirtual reality for rehabilitation and engagement in pediatric intensive care: a systematic review of feasibility, safety and outcomes
Children admitted to pediatric intensive care units (PICUs) are at risk of acquired weakness, delirium, and long-term functional and neurocognitive morbidity. Although early rehabilitation is recommended to mitigate these adverse outcomes, its delivery remains challenging. Virtual reality (VR) has been proposed as a means of enhancing engagement in rehabilitation in this setting. This systematic review evaluated the feasibility, safety, and outcomes of VR delivered to children during routine pediatric intensive care to support rehabilitation, participation in rehabilitation, or emotional regulation. Three databases were searched through 8 August 2026, and the reference list of one relevant scoping review was screened; two reviewers independently screened all records (Cohen's κ = 0.85). Six studies involving 188 children were included: one randomized pilot trial, one prospective single-arm trial, two cross-sectional single-arm studies, and two case reports. Risk of bias was high, serious, or critical in every appraised study, and the certainty of evidence was very low across all outcome domains. The randomized trial found higher consciousness scores at 2 weeks in the VR group but no difference in social and personal functional ability at 3 months. In uncontrolled studies, most children were rated as highly engaged and reported enjoying VR, while parents reported a calming effect more frequently than the children themselves; a transient increase in one heart rate variability parameter occurred during VR but was not sustained afterward. Adverse events were reported in four of the six studies at rates of 3.1% to 20%, comprising dizziness, nausea, visual discomfort, and mild irritability, with one serious event adjudicated as unrelated to VR; the two studies that applied pre-specified criteria recorded no events and one unrelated event, respectively. VR appears feasible and safe for hemodynamically stable, alert children in the PICU and is consistently associated with high engagement. However, whether VR improves functional outcomes remains uncertain. Eligibility, rather than tolerability, may be the principal barrier to implementation, and the available evidence applies to only a minority of the PICU population. Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/view/CRD42024607286 , PROSPERO, CRD42024607286
- 19 June 2026
- Published by American Association for the Advancement of Science (AAAS) in Computational and Structural Biotechnology Journal
- Vol. 35 (1) , 0160
NCES: A Cell-Specific Network-Augmented Essentiality Framework for Cancer Therapeutic Target Discovery
Identifying effective therapeutic targets remains a central challenge in cancer research. CRISPR–Cas9 knockout screens have provided valuable insights into gene essentiality; however, using essentiality at the level of individual genes often fails to reliably distinguish true therapeutic targets from nonfunctional candidates. To address this limitation, we developed the neighbor-correlation essentiality score (NCES), a network-augmented framework that leverages the essentialities of functionally active neighboring genes. NCES combines DepMap CERES scores, which estimate gene essentiality from CRISPR–Cas9 knockout screens, with protein–protein interaction networks. Interaction weights are assigned to network neighbors based on cell-line-specific expression correlations derived from CRISPR knockout or compound-perturbation profiles. The proposed NCES framework was systematically evaluated across 7 cancer cell lines against therapeutic target gold standards. NCES variants consistently outperformed approaches based solely on individual gene essentiality, with the CRISPR-weighted variant achieving the best performance, yielding AUROCs of 0.794 and 0.779 against the Therapeutic Target Database and DrugBank gold standards, respectively. Statistical testing demonstrated that weighted NCES variants significantly improved predictive accuracy over their unweighted counterpart. Finally, several high-ranking genes beyond current gold-standard datasets, including CCNB1 , CDC7 , and WEE1 , were supported by the existing literature as biologically essential or therapeutically actionable. Together, these results demonstrate that NCES advances therapeutic target discovery by leveraging cell-specific, functionally relevant interactions among network neighbors within genome-scale essentiality data.
- 19 June 2026
- Published by American Association for the Advancement of Science (AAAS) in Computational and Structural Biotechnology Journal
- Vol. 35 (1) , 0160
NCES: A Cell-Specific Network-Augmented Essentiality Framework for Cancer Therapeutic Target Discovery
Identifying effective therapeutic targets remains a central challenge in cancer research. CRISPR–Cas9 knockout screens have provided valuable insights into gene essentiality; however, using essentiality at the level of individual genes often fails to reliably distinguish true therapeutic targets from nonfunctional candidates. To address this limitation, we developed the neighbor-correlation essentiality score (NCES), a network-augmented framework that leverages the essentialities of functionally active neighboring genes. NCES combines DepMap CERES scores, which estimate gene essentiality from CRISPR–Cas9 knockout screens, with protein–protein interaction networks. Interaction weights are assigned to network neighbors based on cell-line-specific expression correlations derived from CRISPR knockout or compound-perturbation profiles. The proposed NCES framework was systematically evaluated across 7 cancer cell lines against therapeutic target gold standards. NCES variants consistently outperformed approaches based solely on individual gene essentiality, with the CRISPR-weighted variant achieving the best performance, yielding AUROCs of 0.794 and 0.779 against the Therapeutic Target Database and DrugBank gold standards, respectively. Statistical testing demonstrated that weighted NCES variants significantly improved predictive accuracy over their unweighted counterpart. Finally, several high-ranking genes beyond current gold-standard datasets, including CCNB1 , CDC7 , and WEE1 , were supported by the existing literature as biologically essential or therapeutically actionable. Together, these results demonstrate that NCES advances therapeutic target discovery by leveraging cell-specific, functionally relevant interactions among network neighbors within genome-scale essentiality data.
- Open Access
- 19 May 2026
- Published by Springer Nature in Journal of Cheminformatics
- Vol. 18 (1)
Novel molecular design via a scaffold-aware transformer with multi-scale attention mechanisms
Recent advancements in artificial intelligence have demonstrated enhanced potential in accelerating drug discovery by exploring vast chemical spaces and predicting molecular properties. In particular, scaffold-constrained molecular generation has enabled the incorporation of structural constraints into the generation process. However, effectively integrating multi-scale structural information with activity-guided optimization remains challenging. To overcome these limitations, we propose a novel framework that integrates a transformer-based generative model and a graph attention network-based predictive model. The generative model produces molecules with desired structural characteristics by explicitly incorporating scaffold information, while the predictive model estimates the biological activity of the generated molecules. A supervised fine-tuning framework iteratively refines the generator through multi-stage tournament selection with experience memory. This framework guides the generator toward scaffold-consistent, high-affinity candidates while exploring novel chemical variations around a user-specified scaffold. Experimental results demonstrated that the proposed scaffold-aware transformer achieves competitive validity, uniqueness, and novelty, effectively generating novel compounds with high predicted binding affinity for biological targets. Meanwhile, an attention-based analysis extracted atom-level importance scores, highlighting the substructures that contribute to the predicted binding affinity and providing interpretable insights into structure–activity relationships. This study provides a practical and interpretable tool for scaffold-conditioned molecular generation.
- 19 May 2026
- Published by Springer Nature in Journal of Cheminformatics
- Vol. 18 (1)
Open AccessNovel molecular design via a scaffold-aware transformer with multi-scale attention mechanisms
Recent advancements in artificial intelligence have demonstrated enhanced potential in accelerating drug discovery by exploring vast chemical spaces and predicting molecular properties. In particular, scaffold-constrained molecular generation has enabled the incorporation of structural constraints into the generation process. However, effectively integrating multi-scale structural information with activity-guided optimization remains challenging. To overcome these limitations, we propose a novel framework that integrates a transformer-based generative model and a graph attention network-based predictive model. The generative model produces molecules with desired structural characteristics by explicitly incorporating scaffold information, while the predictive model estimates the biological activity of the generated molecules. A supervised fine-tuning framework iteratively refines the generator through multi-stage tournament selection with experience memory. This framework guides the generator toward scaffold-consistent, high-affinity candidates while exploring novel chemical variations around a user-specified scaffold. Experimental results demonstrated that the proposed scaffold-aware transformer achieves competitive validity, uniqueness, and novelty, effectively generating novel compounds with high predicted binding affinity for biological targets. Meanwhile, an attention-based analysis extracted atom-level importance scores, highlighting the substructures that contribute to the predicted binding affinity and providing interpretable insights into structure–activity relationships. This study provides a practical and interpretable tool for scaffold-conditioned molecular generation.
- Open Access
- 4 May 2026
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 27 (3)
Carcinogenicity prediction via multi-task learning of cross-organ representations with attention mechanisms
Cancer is caused by the uncontrolled growth and division of abnormal cells. In industrialized societies, chemical exposure is a leading cause of cancer. Since certain compounds induce cancer by damaging genes or affecting cellular metabolism, studying carcinogens is essential. However, previous studies used separate models for each organ and failed to capture carcinogenic features shared across organs, limiting generalization. Thus, this study developed a multi-task learning framework to predict organ-specific carcinogenicity in the liver, lung, stomach, and breast. This framework consisted of a shared layer and task-specific layers. The shared layer contains a graph attention network layer to make atom-level representations, along with parallel fully connected layers designed for each task combination. The resulting shared representations are passed to task-specific layers to predict organ-specific carcinogenicity. The training process followed stepwise learning, whereby the model was first trained using partially labeled data to capture cross-organ representations and determine initial weights. In the second step, fully labeled data for all organs were used for final training. The proposed multi-task model achieved superior performance in the liver, lung, and stomach tasks. Notably, it recorded the highest area under the receiver operating characteristic curve in the stomach task (0.7636), outperforming the single-task model (0.7055) and all comparative models (0.5527–0.7418). The highest area under the precision–recall curve was observed in the liver task (0.9646), surpassing the single-task model (0.9505) and all comparative models (0.9373–0.9621). We further analyzed molecules with high predicted carcinogenicity and identified critical substructures using an attention mechanism. This research can contribute to predicting organ-specific carcinogenicity of candidate chemicals in the early stages of drug development.
- 4 May 2026
- Published by Oxford University Press (OUP) in Briefings in Bioinformatics
- Vol. 27 (3)
Open AccessCarcinogenicity prediction via multi-task learning of cross-organ representations with attention mechanisms
Cancer is caused by the uncontrolled growth and division of abnormal cells. In industrialized societies, chemical exposure is a leading cause of cancer. Since certain compounds induce cancer by damaging genes or affecting cellular metabolism, studying carcinogens is essential. However, previous studies used separate models for each organ and failed to capture carcinogenic features shared across organs, limiting generalization. Thus, this study developed a multi-task learning framework to predict organ-specific carcinogenicity in the liver, lung, stomach, and breast. This framework consisted of a shared layer and task-specific layers. The shared layer contains a graph attention network layer to make atom-level representations, along with parallel fully connected layers designed for each task combination. The resulting shared representations are passed to task-specific layers to predict organ-specific carcinogenicity. The training process followed stepwise learning, whereby the model was first trained using partially labeled data to capture cross-organ representations and determine initial weights. In the second step, fully labeled data for all organs were used for final training. The proposed multi-task model achieved superior performance in the liver, lung, and stomach tasks. Notably, it recorded the highest area under the receiver operating characteristic curve in the stomach task (0.7636), outperforming the single-task model (0.7055) and all comparative models (0.5527–0.7418). The highest area under the precision–recall curve was observed in the liver task (0.9646), surpassing the single-task model (0.9505) and all comparative models (0.9373–0.9621). We further analyzed molecules with high predicted carcinogenicity and identified critical substructures using an attention mechanism. This research can contribute to predicting organ-specific carcinogenicity of candidate chemicals in the early stages of drug development.
- Open Access
- 23 April 2026
- Published by MDPI AG in Electronics
- Vol. 15 (9) , 1798
End-to-End Learnable Recurrence Plot for Sleep Stage Classification Using Non-Contact Ballistocardiography
Accurate sleep stage classification is essential for evaluating sleep quality, yet clinical polysomnography is impractical for continuous home-based monitoring. Ballistocardiography (BCG) enables unobtrusive sleep monitoring through sensors embedded in sleep furniture; however, existing BCG-based approaches either rely on complex physiological feature extraction or employ fixed-parameter signal-to-image transformations that cannot adapt to inter-subject variability. This study proposes a learnable recurrence plot (RP) framework for three-stage sleep classification (Wake, NREM, REM) from single-channel BCG signals. The Learnable RP introduces three innovations: multi-scale phase-space reconstruction at physiologically motivated time delays (τ = 5, 10, 20), differentiable per-scale thresholds optimized end-to-end, and attention-based spatial fusion of multi-scale recurrence maps. The framework was evaluated through 10-fold stratified cross-validation across six backbone architectures using 50 overnight recordings. The Learnable RP consistently outperformed four baseline transformation methods (GAF, MTF, Classical RP, Modified RP), achieving an aggregate mean accuracy of 73.60%, with EfficientNet-B5 reaching 78.91%. and 78.91%. Statistical validation across all 24 pairwise comparisons (4 baselines × 6 backbones) confirmed consistent superiority (all p < 0.001). The proposed framework achieves competitive performance without explicit physiological feature engineering, offering a viable path toward end-to-end unobtrusive sleep monitoring.
- 23 April 2026
- Published by MDPI AG in Electronics
- Vol. 15 (9) , 1798
Open AccessEnd-to-End Learnable Recurrence Plot for Sleep Stage Classification Using Non-Contact Ballistocardiography
Accurate sleep stage classification is essential for evaluating sleep quality, yet clinical polysomnography is impractical for continuous home-based monitoring. Ballistocardiography (BCG) enables unobtrusive sleep monitoring through sensors embedded in sleep furniture; however, existing BCG-based approaches either rely on complex physiological feature extraction or employ fixed-parameter signal-to-image transformations that cannot adapt to inter-subject variability. This study proposes a learnable recurrence plot (RP) framework for three-stage sleep classification (Wake, NREM, REM) from single-channel BCG signals. The Learnable RP introduces three innovations: multi-scale phase-space reconstruction at physiologically motivated time delays (τ = 5, 10, 20), differentiable per-scale thresholds optimized end-to-end, and attention-based spatial fusion of multi-scale recurrence maps. The framework was evaluated through 10-fold stratified cross-validation across six backbone architectures using 50 overnight recordings. The Learnable RP consistently outperformed four baseline transformation methods (GAF, MTF, Classical RP, Modified RP), achieving an aggregate mean accuracy of 73.60%, with EfficientNet-B5 reaching 78.91%. and 78.91%. Statistical validation across all 24 pairwise comparisons (4 baselines × 6 backbones) confirmed consistent superiority (all p < 0.001). The proposed framework achieves competitive performance without explicit physiological feature engineering, offering a viable path toward end-to-end unobtrusive sleep monitoring.
- Open Access
- 21 April 2026
- Published by JMIR Publications Inc. in JMIR Public Health and Surveillance
- Vol. 12 (1) , e88408
Urinary Sodium Excretion and the Risk of Prevalent Anemia: Nationwide Population-Based Cross-Sectional Study
Background: While excessive dietary sodium intake is an established risk factor for cardiovascular and renal complications, its potential association with anemia remains largely unexplored. Objective: We hypothesized that, based on the observed benefits of sodium-glucose cotransporter 2 inhibitors in correcting anemia through alterations in renal tubular metabolism and oxygen homeostasis, elevated urinary sodium excretion, as a surrogate of dietary sodium intake, may increase the risk of prevalent anemia. Methods: This nationwide cross-sectional study analyzed 54,802 adults from the Korea National Health and Nutrition Examination Survey (2014-2023). Participants were stratified by spot urine sodium-to-creatinine ratio (Na+/Cr) quartiles (first quartile [Q1], second quartile [Q2], third quartile [Q3], and fourth quartile [Q4]). Anemia was defined as hemoglobin <13 g/dL for men and <12 g/dL for women. Results: Anemia prevalence increased progressively across spot urine Na+/Cr quartiles (Q1: 925/13,700, 6.8%, Q2: 1126/13,701, 8.2%, Q3: 1393/13,701, 10.2%, and Q4: 1893/13,700, 13.8%). Multivariable logistic regression demonstrated that participants in the highest quartile had 43% higher odds of anemia compared with the lowest quartile (adjusted odds ratio 1.429, 95% CI 1.269-1.610; P<.001). Each 1-unit log increase in spot urine Na+/Cr conferred a 67% increase in odds of anemia (adjusted odds ratio 1.674, 95% CI 1.452-1.930; P<.001). Sensitivity analyses using tertiles, quintiles, estimated 24-hour sodium excretion, and restriction to preserved kidney function consistently confirmed these associations. Conclusions: Higher urinary sodium excretion exhibits a robust, graded association with increased anemia prevalence in the general population. These findings suggest that dietary sodium restriction may provide additional benefits beyond cardiovascular protection.
- 21 April 2026
- Published by JMIR Publications Inc. in JMIR Public Health and Surveillance
- Vol. 12 (1) , e88408
Open AccessUrinary Sodium Excretion and the Risk of Prevalent Anemia: Nationwide Population-Based Cross-Sectional Study
Background: While excessive dietary sodium intake is an established risk factor for cardiovascular and renal complications, its potential association with anemia remains largely unexplored. Objective: We hypothesized that, based on the observed benefits of sodium-glucose cotransporter 2 inhibitors in correcting anemia through alterations in renal tubular metabolism and oxygen homeostasis, elevated urinary sodium excretion, as a surrogate of dietary sodium intake, may increase the risk of prevalent anemia. Methods: This nationwide cross-sectional study analyzed 54,802 adults from the Korea National Health and Nutrition Examination Survey (2014-2023). Participants were stratified by spot urine sodium-to-creatinine ratio (Na+/Cr) quartiles (first quartile [Q1], second quartile [Q2], third quartile [Q3], and fourth quartile [Q4]). Anemia was defined as hemoglobin <13 g/dL for men and <12 g/dL for women. Results: Anemia prevalence increased progressively across spot urine Na+/Cr quartiles (Q1: 925/13,700, 6.8%, Q2: 1126/13,701, 8.2%, Q3: 1393/13,701, 10.2%, and Q4: 1893/13,700, 13.8%). Multivariable logistic regression demonstrated that participants in the highest quartile had 43% higher odds of anemia compared with the lowest quartile (adjusted odds ratio 1.429, 95% CI 1.269-1.610; P<.001). Each 1-unit log increase in spot urine Na+/Cr conferred a 67% increase in odds of anemia (adjusted odds ratio 1.674, 95% CI 1.452-1.930; P<.001). Sensitivity analyses using tertiles, quintiles, estimated 24-hour sodium excretion, and restriction to preserved kidney function consistently confirmed these associations. Conclusions: Higher urinary sodium excretion exhibits a robust, graded association with increased anemia prevalence in the general population. These findings suggest that dietary sodium restriction may provide additional benefits beyond cardiovascular protection.
- Open Access
- 7 April 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (4)
Predicting condition-aware drug-induced transcriptional responses via a latent diffusion model
Motivation: Accurate prediction of condition-aware drug-induced transcriptional responses is essential for drug discovery and precision medicine. Current computational models, including encoder–decoder architectures and generative adversarial network-based approaches, exhibit reasonable performance but frequently neglect biological characteristics and fail to generalize to unseen conditions. Thus, this study presents a latent diffusion model that combines a variational autoencoder (VAE) with a diffusion process. Results: The VAE compresses gene expression (GE) profiles into a low-dimensional latent space, where the diffusion process learns the joint probability distribution over latent GE representations and noisy intermediates, thereby enabling more effective capture of gene–gene correlations. The model incorporates multiple perturbation conditions, including cell line, compound, dose, and time, to enhance prediction performance on unseen conditions. The reverse diffusion process predicts both the mean and variance of the posterior distribution, improving the fidelity of the generated GE profiles. The proposed model achieved the highest reconstruction performance in the unseen compound split with a Pearson correlation coefficient of 0.870 ± 0.001 and an R2 score of 0.739 ± 0.001, outperforming previous approaches. The model demonstrated superior preservation of gene–gene correlations, as confirmed by heatmap analysis. To evaluate biological relevance, we predicted half-maximal inhibitory concentration using generated GE, outperforming baseline methods. Latent space analysis revealed that the model preserved cell line identity and continuous dose–time variation. Gene set enrichment analysis confirmed that predicted GE reproduced known pathway-level responses to perturbation. These results demonstrate diffusion-based generative models as effective tools for modeling transcriptional responses in drug discovery and precision medicine. Availability and implementation: Source code and dataset are available at https://doi.org/10.5281/zenodo.18871024.
- 7 April 2026
- Published by Oxford University Press (OUP) in Bioinformatics
- Vol. 42 (4)
Open AccessPredicting condition-aware drug-induced transcriptional responses via a latent diffusion model
Motivation: Accurate prediction of condition-aware drug-induced transcriptional responses is essential for drug discovery and precision medicine. Current computational models, including encoder–decoder architectures and generative adversarial network-based approaches, exhibit reasonable performance but frequently neglect biological characteristics and fail to generalize to unseen conditions. Thus, this study presents a latent diffusion model that combines a variational autoencoder (VAE) with a diffusion process. Results: The VAE compresses gene expression (GE) profiles into a low-dimensional latent space, where the diffusion process learns the joint probability distribution over latent GE representations and noisy intermediates, thereby enabling more effective capture of gene–gene correlations. The model incorporates multiple perturbation conditions, including cell line, compound, dose, and time, to enhance prediction performance on unseen conditions. The reverse diffusion process predicts both the mean and variance of the posterior distribution, improving the fidelity of the generated GE profiles. The proposed model achieved the highest reconstruction performance in the unseen compound split with a Pearson correlation coefficient of 0.870 ± 0.001 and an R2 score of 0.739 ± 0.001, outperforming previous approaches. The model demonstrated superior preservation of gene–gene correlations, as confirmed by heatmap analysis. To evaluate biological relevance, we predicted half-maximal inhibitory concentration using generated GE, outperforming baseline methods. Latent space analysis revealed that the model preserved cell line identity and continuous dose–time variation. Gene set enrichment analysis confirmed that predicted GE reproduced known pathway-level responses to perturbation. These results demonstrate diffusion-based generative models as effective tools for modeling transcriptional responses in drug discovery and precision medicine. Availability and implementation: Source code and dataset are available at https://doi.org/10.5281/zenodo.18871024.
- Open Access
- 31 March 2026
- Published by Digital Contents Society in Journal of Digital Contents Society
- Vol. 27 (3) , 831-838
Predicting Synergistic Anticancer Drug Combinations via Graph Attention Network
- 31 March 2026
- Published by Digital Contents Society in Journal of Digital Contents Society
- Vol. 27 (3) , 831-838
Open AccessPredicting Synergistic Anticancer Drug Combinations via Graph Attention Network
- 6 March 2026
- Published by Elsevier in Food and Chemical Toxicology
- Vol. 212, 116050
Combining UHPLC profiling and random walk network-based in vitro analysis to predict herb-induced liver injury
- 6 March 2026
- Published by Elsevier in Food and Chemical Toxicology
- Vol. 212, 116050
Combining UHPLC profiling and random walk network-based in vitro analysis to predict herb-induced liver injury