Bioinformatics. 2025 Apr 11:btaf160. doi: 10.1093/bioinformatics/btaf160. Online ahead of print.

ABSTRACT

MOTIVATION: Leveraging deep learning for the representation learning of Gene Ontology (GO) and Gene Ontology Annotation (GOA) holds significant promise for enhancing downstream biological tasks such as protein-protein interaction prediction. Prior approaches have predominantly used text- and graph-based methods, embedding GO and GOA in a single geometric space (e.g., Euclidean or hyperbolic). However, since the GO graph exhibits a complex and non-monotonic hierarchy, single-space embeddings are insufficient to fully capture its structural nuances.

RESULTS: In this study, we address this limitation by exploiting geometric interaction to better reflect the intricate hierarchical structure of GO. Our proposed method, Geometry-Aware Knowledge Graph Embeddings for GO and Genes (GeOKG), leverages interactions among various geometric representations during training, thereby modeling the complex hierarchy of GO more effectively. Experiments at the GO level demonstrate the benefits of incorporating these geometric interactions, while gene-level tests reveal that GeOKG outperforms existing methods in protein-protein interaction prediction. These findings highlight the potential of using geometric interaction for embedding heterogeneous biomedical networks.

AVAILABILITY AND IMPLEMENTATION: https://github.com/ukjung21/GeOKG.

SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

PMID:40217132 | DOI:10.1093/bioinformatics/btaf160

Sci Rep. 2025 Apr 11;15(1):12482. doi: 10.1038/s41598-025-97565-4.

ABSTRACT

Ensuring data privacy in medical image classification is a critical challenge in healthcare, especially with the increasing reliance on AI-driven diagnostics. In fact, over 30% of healthcare organizations globally have experienced a data breach in the last year, highlighting the need for secure solutions. This study investigates the integration of transfer learning and federated learning for privacy-preserving medical image classification using GoogLeNet and VGG16 as baseline models to evaluate the generalizability of the proposed framework. Pre-trained on ImageNet and fine-tuned on three specialized medical datasets for TB chest X-rays, brain tumor MRI scans, and diabetic retinopathy images, these models achieved high classification accuracy across various aggregation methods. Additionally, the proposed dynamic aggregation method was further analyzed using modern architectures, EfficientNetV2 and ResNet-RS, to assess the scalability and robustness of the model. A key contribution is the introduction of a novel adaptive aggregation method, which dynamically alternates between Federated Averaging (FedAvg) and Federated Stochastic Gradient Descent (FedSGD), based on data divergence during communication rounds. This approach optimizes model convergence while preserving privacy in collaborative settings. The results demonstrate that transfer learning, when combined with federated learning, offers a scalable, robust, and secure solution for real-world medical diagnostics, enabling healthcare institutions to train highly accurate models without compromising sensitive patient data.

PMID:40217112 | DOI:10.1038/s41598-025-97565-4

Sci Rep. 2025 Apr 11;15(1):12383. doi: 10.1038/s41598-025-97052-w.

ABSTRACT

Hyperspectral imaging acquired from unmanned aerial vehicles (UAVs) offers detailed spectral and spatial data that holds transformative potential for precision agriculture applications, such as crop classification, health monitoring, and yield estimation. However, traditional methods struggle to effectively capture both local and global features, particularly in complex agricultural environments with diverse crop types, varying growth stages, and imbalanced data distributions. To address these challenges, we propose CMTNet, an innovative deep learning framework that integrates convolutional neural networks (CNNs) and Transformers for hyperspectral crop classification. The model combines a spectral-spatial feature extraction module to capture shallow features, a dual-branch architecture that extracts both local and global features simultaneously, and a multi-output constraint module to enhance classification accuracy through cross-constraints among multiple feature levels. Extensive experiments were conducted on three UAV-acquired datasets: WHU-Hi-LongKou, WHU-Hi-HanChuan, and WHU-Hi-HongHu. The experimental results demonstrate that CMTNet achieved overall accuracy (OA) values of 99.58%, 97.29%, and 98.31% on these three datasets, surpassing the current state-of-the-art method (CTMixer) by 0.19% (LongKou), 1.75% (HanChuan), and 2.52% (HongHu) in OA values, respectively. These findings indicate its superior potential for UAV-based agricultural monitoring in complex environments. These results advance the precision and reliability of hyperspectral crop classification, offering a valuable solution for precision agriculture challenges.

PMID:40216979 | DOI:10.1038/s41598-025-97052-w

NPJ Digit Med. 2025 Apr 11;8(1):204. doi: 10.1038/s41746-025-01606-1.

ABSTRACT

The novel coronavirus (COVID-19) pandemic has had a devastating global impact, profoundly affecting daily life, healthcare systems, and public health infrastructure. Despite the availability of treatments and vaccines, hospitalizations and deaths continue. Real-time surveillance of infection trends supports resource allocation and mitigation strategies, but reliable forecasting remains a challenge. While deep learning has advanced time-series forecasting, its effectiveness relies on large datasets, a significant obstacle given the pandemic's evolving nature. Most models use national or state-level data, limiting both dataset size and the granularity of insights. To address this, we propose the Fine-Grained Infection Forecast Network (FIGI-Net), a stacked bidirectional LSTM structure designed to leverage county-level data to produce daily forecasts up to two weeks in advance. FIGI-Net outperforms existing models, accurately predicting sudden changes such as new outbreaks or peaks, a capability many state-of-the-art models lack. This approach could enhance public health responses and outbreak preparedness.

PMID:40216974 | DOI:10.1038/s41746-025-01606-1

Sci Rep. 2025 Apr 11;15(1):12453. doi: 10.1038/s41598-025-97564-5.

ABSTRACT

Lymphedema is a chronic condition characterized by lymphatic fluid accumulation, primarily affecting the limbs. Its diagnosis is challenging due to symptom overlap with conditions like chronic venous insufficiency (CVI), deep vein thrombosis (DVT), and systemic diseases, often leading to diagnostic delays that can extend up to ten years. These delays negatively impact patient outcomes and burden healthcare systems. Conventional diagnostic methods rely heavily on clinical expertise, which may fail to distinguish subtle variations between these conditions. This study investigates the application of artificial intelligence (AI), specifically deep learning, to improve diagnostic accuracy for lower limb edema. A dataset of 1622 clinical images was used to train sixteen convolutional neural networks (CNNs) and transformer-based models, including EfficientNetV2, which achieved the highest accuracy of 78.6%. Grad-CAM analyses enhanced model interpretability, highlighting clinically relevant features such as swelling and hyperpigmentation. The AI system consistently outperformed human evaluators, whose diagnostic accuracy plateaued at 62.7%. The findings underscore the transformative potential of AI as a diagnostic tool, particularly in distinguishing conditions with overlapping clinical presentations. By integrating AI with clinical workflows, healthcare systems can reduce diagnostic delays, enhance accuracy, and alleviate the burden on medical professionals. While promising, the study acknowledges limitations, such as dataset diversity and the controlled evaluation environment, which necessitate further validation in real-world settings. This research highlights the potential of AI-driven diagnostics to revolutionize lymphedema care, bridging gaps in conventional methods and supporting healthcare professionals in delivering more precise and timely interventions. Future work should focus on external validation and hybrid systems integrating AI and clinical expertise for comprehensive diagnostic solutions.

PMID:40216943 | DOI:10.1038/s41598-025-97564-5

NPJ Digit Med. 2025 Apr 11;8(1):203. doi: 10.1038/s41746-025-01607-0.

ABSTRACT

Traditional sleep staging categorizes sleep and wakefulness into five coarse-grained classes, overlooking subtle variations within each stage. We propose a deep learning method to annotate continuous sleep depth index (SDI) with existing discrete sleep staging labels, using polysomnography from over 10,000 recordings across four large-scale cohorts. The results showcased a strong correlation between the decrease in sleep depth index and the increase in duration of arousal. Case studies indicated that SDI captured more nuanced sleep structures than conventional sleep staging. Clustering based on the digital biomarkers extracted from the SDI identified two subtypes of sleep, where participants in the disturbed subtype had a higher prevalence of several poor health conditions and were associated with a 33% increased risk of mortality and a 38% increased risk of fatal coronary heart disease. Our study underscores the utility of SDI in revealing more detailed sleep structures and yielding novel digital biomarkers for sleep medicine.

PMID:40216900 | DOI:10.1038/s41746-025-01607-0

Sci Rep. 2025 Apr 11;15(1):12442. doi: 10.1038/s41598-025-95307-0.

ABSTRACT

Diffuse pulmonary ossification (DiPO) is characterised by widespread ectopic bone formation in the lungs. Idiopathic DiPO (I-DiPO) poses significant diagnostic challenges and its ossification mechanism remains unclear. Cicatricial organising pneumonia (CiOP) lesions form fibrous nodules without damaging lung structure. We investigated the histopathological features of I-DiPO, focusing on the surrounding fibrosis, and compared them with those of secondary DiPO (S-DiPO). An analysis was conducted using data from a nationwide DiPO survey in Japan. The dataset included clinical, radiological, and histopathological data of patients with suspected I-DiPO. The specific patterns of ossification and fibrotic findings such as CiOP, organising pneumonia (OP), and subpleural fibrosis were identified. Eighteen and seven patients were classified as having I-DiPO and S-DiPO, respectively. I-DiPO affects younger patients, progresses slowly, commonly occurs in the lower lungs, and has a lower mortality rate. S-DiPO affects older patients, presents with widespread lung lesions, and has a higher mortality rate. CiOP lesions were found in direct continuity with or near ossified lesions in 61.1% and 71.4% of patients with I-DiPO and S-DiPO, respectively. OP, CiOP, and ossified lesions often observed in the same locations in S-DiPO. DiPO has a unique pathogenesis, with an ossification transition occurring via the CiOP lesions. These findings provide valuable insights for future diagnostic approaches and management strategies for this condition.

PMID:40216850 | DOI:10.1038/s41598-025-95307-0

Cell Death Dis. 2025 Apr 11;16(1):278. doi: 10.1038/s41419-025-07573-5.

ABSTRACT

Idiopathic pulmonary fibrosis has a poor prognosis, with existing medications only partially alleviating symptoms, highlighting the urgent need for new therapeutic approaches. The dysregulations of Rho GTPases/ROCK are related with various diseases, including fibrosis. Nevertheless, the development of drugs for pulmonary fibrosis treatment has predominantly concentrated on ROCK inhibitors. Small GTPases have been historically recognized as "undruggable". Here, we explore a novel Rho GEFs inhibitor GL-V9, and find that GL-V9 alleviates bleomycin-induced pulmonary fibrosis in mice by inhibiting myofibroblast activation and reprogramming profibrotic macrophages. Distinct from the mechanisms of the first-line drug Nintedanib, GL-V9 binds to the DH/PH domain of Rho GEFs and block the activation of Rho GTPase signaling. This action subsequently suppresses myofibroblast activation by interfering with Rho GTPase-dependent cytoskeletal reorganization and the activity of MRTF and YAP, and inhibits M2 macrophage polarization by modulating RhoA/STAT3 activity. The discovery of new regulatory mechanisms of GL-V9 suggests that targeting Rho GEFs represents a potent strategy for pulmonary fibrosis treatment.

PMID:40216763 | DOI:10.1038/s41419-025-07573-5

Gut Pathog. 2025 Apr 11;17(1):23. doi: 10.1186/s13099-025-00689-1.

ABSTRACT

BACKGROUND: Breastfeeding is a major determinant of gut microbiota composition and fermentation activity during the first months of life. Breastmilk delivers human milk oligosaccharides (HMO) as substrates for microbial intestinal fermentation. One of the main metabolites that accumulates in feces of breastfed infants is 1,2-propanediol (1,2PD) resulting from the metabolism of fucosylated HMO. 1,2PD is used in microbial cross-feeding to produce propionate, but 1,2PD is also an alcohol that can impact the state of the microbial cell envelope. To shed further light on an understudied compound in the infant gut, we investigated the genetic and metabolic potential of the early gut colonizer Clostridium perfringens to utilise 1,2PD, and the interactions of 1,2PD with the cell envelope.

RESULTS: Based on genome analysis, C. perfringens FMT 1006 isolated from infant feces possessed most genes of the pdu operon related to 1,2PD metabolism. C. perfringens consumed 1,2PD (78%) and produced 1-propanol as the main metabolite, while propionate was not detected. In agreement, genes responsible for 1,2PD utilisation and propanol formation (pduCDE, dhaT) were highly expressed. When cultivated in the presence of 1,2PD and glucose, a higher proportion of 1,2PD carbon (87%) was recovered as compared to incubation with only 1,2PD (34%). At the same time, lactate and acetate were formed in a ratio of 2.16:1.0 with 1,2PD and glucose compared to a ratio 9.0:1.0 during growth with only glucose possibly due to reallocation of the NAD+/NADH pool in favor of 1-propanol formation. The presence of 1,2PD slightly increased membrane fluidity and modified the composition of the membrane to a higher content of elongated glycerophosphoethanolamines.

CONCLUSION: We provide here new knowledge on the metabolism of 1,2PD by a microbial species that is present during breastfeeding and observed that C. perfringens metabolised 1,2PD mainly to propanol. The presence of 1,2PD had little impact on membrane fluidity and let to modifications of membrane lipid composition. Collectively, these findings advance our understanding of on intestinal metabolite-microbe interactions during breastfeeding.

PMID:40217307 | DOI:10.1186/s13099-025-00689-1

Nat Microbiol. 2025 Apr 11. doi: 10.1038/s41564-025-01972-2. Online ahead of print.

ABSTRACT

Metabolic models of individual microorganisms or small microbial consortia have become standard research tools in the bioengineering and systems biology fields. However, extending metabolic modelling to diverse microbial communities, such as those in the human gut, remains a practical challenge from both modelling and experimental validation perspectives. In complex communities, metabolic models accounting for community dynamics, or those that consider multiple objectives, may provide optimal predictions over simpler steady-state models, but require a much higher computational cost. Here we describe some of the strengths and limitations of microbial community-scale metabolic models and argue for a robust validation framework for developing personalized, mechanistic and accurate predictions of microbial community metabolic behaviours across environmental contexts. Ultimately, quantitatively accurate microbial community-scale metabolic models could aid in the design and testing of personalized prebiotic, probiotic and dietary interventions that optimize for translationally relevant outcomes.

PMID:40217129 | DOI:10.1038/s41564-025-01972-2

Nat Med. 2025 Apr 11. doi: 10.1038/s41591-025-03615-9. Online ahead of print.

ABSTRACT

The landscape of clinical microbiome research has dramatically evolved over the past decade. By leveraging in vivo and in vitro experimentation, multiomic approaches and computational biology, we have uncovered mechanisms of action and microbial metrics of association and identified effective ways to modify the microbiome in many diseases and treatment modalities. This Review explores recent advances in the clinical application of microbiome research over the past 5 years, while acknowledging existing barriers and highlighting opportunities. We focus on the translation of microbiome research into clinical practice, spearheaded by Food and Drug Administration (FDA)-approved microbiome therapies for recurrent Clostridioides difficile infections and the emerging fields of microbiome-based diagnostics and therapeutics. We highlight key examples of studies demonstrating how microbiome mechanisms, metrics and modifiers can advance clinical practice. We also discuss forward-looking perspectives on key challenges and opportunities toward integrating microbiome data into routine clinical practice, precision medicine and personalized healthcare and nutrition.

PMID:40217076 | DOI:10.1038/s41591-025-03615-9

Oncogene. 2025 Apr 11. doi: 10.1038/s41388-025-03362-8. Online ahead of print.

NO ABSTRACT

PMID:40216970 | DOI:10.1038/s41388-025-03362-8

Commun Chem. 2025 Apr 11;8(1):114. doi: 10.1038/s42004-025-01484-4.

ABSTRACT

Computational techniques for predicting molecular properties are emerging as key components for streamlining drug development, optimizing time and financial investments. Here, we introduce ChemLM, a transformer language model for this task. ChemLM leverages self-supervised domain adaptation on chemical molecules to enhance its predictive performance. Within the framework of ChemLM, chemical compounds are conceptualized as sentences composed of distinct chemical 'words', which are employed for training a specialized chemical language model. On the standard benchmark datasets, ChemLM either matched or surpassed the performance of current state-of-the-art methods. Furthermore, we evaluated the effectiveness of ChemLM in identifying highly potent pathoblockers targeting Pseudomonas aeruginosa (PA), a pathogen that has shown an increased prevalence of multidrug-resistant strains and has been identified as a critical priority for the development of new medications. ChemLM demonstrated substantially higher accuracy in identifying highly potent pathoblockers against PA when compared to state-of-the-art approaches. An intrinsic evaluation demonstrated the consistency of the chemical language model's representation concerning chemical properties. The results from benchmarking, experimental data and intrinsic analysis of the ChemLM space confirm the wide applicability of ChemLM for enhancing molecular property prediction within the chemical domain.

PMID:40216964 | DOI:10.1038/s42004-025-01484-4

Metabolomics. 2025 Apr 11;21(3):51. doi: 10.1007/s11306-025-02251-1.

NO ABSTRACT

PMID:40216608 | DOI:10.1007/s11306-025-02251-1

Clin Microbiol Infect. 2025 Apr 9:S1198-743X(25)00163-6. doi: 10.1016/j.cmi.2025.04.003. Online ahead of print.

ABSTRACT

BACKGROUND: Candida auris is a multidrug-resistant yeast that emerged as a significant healthcare-associated pathogen. It is classified as an urgent threat to public health due to the high resistance to available antifungal agents. Globally six distinct clades of C. auris have been identified with varying antifungal susceptibility patterns and geographical distributions.

OBJECTIVES: The aim of this review is investigating the (published) antifungal susceptibility profiles of different C. auris clades to identify those with a higher prevalence of resistance.

SOURCES: A comprehensive literature review was conducted using PubMed, SciELO, Google Scholar, and MEDLINE databases to collect data on Minimum Inhibitory Concentration (MIC) distributions and clade designations of C. auris strains.

CONTENT: A total of 1,031 C. auris strains were included. Clades I and III, which are closely related phylogenetically, displayed the highest resistance rates, particularly to fluconazole, with 94% and 96% of isolates, respectively. Clade IV also exhibited resistance to both azoles and echinocandins. In contrast, Clades II, V, and VI had lower resistance rates, with Clade VI being entirely susceptible to fluconazole. Anidulafungin demonstrated the greatest efficacy across all clades, with resistance rates ranging from 0% to 3.67%. Furthermore, Clades V and VI showed complete susceptibility to all antifungal agents evaluated.

IMPLICATIONS: This study highlights significant variations in antifungal resistance profiles across the six C. auris clades. Clades I, III, and IV stand out due to their multidrug resistance, particularly to fluconazole and amphotericin B, posing serious challenges for treatment. Continuous global surveillance and tailored management strategies are essential for controlling C. auris infections, especially in highly resistant clades. Enhanced diagnostic capabilities and further genomic studies are critical to understanding the evolving nature of resistance in this emerging pathogen and improving therapeutic outcomes. Clade-specific antifungal resistance in C. auris requires monitoring to optimize therapy selection during outbreaks.

PMID:40216246 | DOI:10.1016/j.cmi.2025.04.003

J Invest Dermatol. 2025 Apr 9:S0022-202X(25)00401-4. doi: 10.1016/j.jid.2025.03.033. Online ahead of print.

ABSTRACT

T cells and structural cells coordinate appropriate inflammatory responses and restoration of barrier integrity following insult. Dysfunctional T cells precipitate skin pathology occurring alongside altered structural cell frequencies and transcriptional states, but to what extent different T cells promote disease-associated changes remains unclear. We show that functionally diverse circulating and skin-resident CD4+CLA+ T cell populations promote distinct transcriptional outcomes in human keratinocytes and fibroblasts associated with inflamed or healthy tissue. We identify Th17 cell-induced genes in keratinocytes that are enriched in psoriasis patient skin and normalized by anti-IL-17 therapy. We also describe a CD103+ skin-resident T cell-induced transcriptional module enriched in healthy controls that is diminished during psoriasis and scleroderma and show that CD103+ T cell frequencies are altered during disease. Interrogating clinical data using immune-dependent transcriptional signatures defines the T cell subsets and genes distinguishing inflamed from healthy skin and allows investigation of heterogeneous patient responses to biologic therapy.

PMID:40216155 | DOI:10.1016/j.jid.2025.03.033

Int J Biol Macromol. 2025 Apr 9:143040. doi: 10.1016/j.ijbiomac.2025.143040. Online ahead of print.

ABSTRACT

Poly-3-hydroxyalkanoate (PHA) and polyacrylamide (PAM) are potentially copolymerized into a novel hybrid material with distinctive characteristics but scarcely explored. In this study, the copolymerization of semicrystalline and amorphous forms of PHA, designated as scPHA and amPHA, with PAM utilizing ultrasonication and hydrogen peroxide as the initiator under defined conditions were investigated. The effect of varying acrylamide amounts on the yield and properties of graft copolymers (PHA-g-MA) were characterized by molecular weight changes, thermal and spectroscopic properties. Grafting scPHA and amPHA with PAM influenced their initial molecular weights (Mw). Specifically, scPHA's Mw decreased from 140 × 103 to ~130 × 103 g mol-1, while amPHA's Mw increased from 62 × 103 to ~68 × 103 g mol-1. Additionally, scPHA and amPHA copolymers showed an increase in thermal decomposition temperature (Td) from 240 °C to 265 °C and 275 °C to 285 °C, respectively. The presence of an amide functional group in the copolymers was authenticated by a Raman peak at 1100 cm-1. Both scPHA and amPHA graft copolymers exhibited unique contrast to their neat counterpart. The scPHA-g-PAMs showed strong crystalline characteristics, evidenced by elevated glass transition and melting temperatures, alterations in crystalline planes and surface morphologies, and a decrease in dielectric constant values. Conversely, amPHA-g-PAMs showed pronounced amorphous characteristics, substantiated by a lowered glass transition and melting temperatures, altered surface morphologies, and increased dielectric constant values. The observations made are attributed to distinct chain packing characteristics between both graft copolymers, which are the consequence of the PHA alkyl side chains length. The biological PHA graft copolymers may offer a diverse array of applications such composite for tissue engineering and biosensor development.

PMID:40216121 | DOI:10.1016/j.ijbiomac.2025.143040

Cancer Cell. 2025 Apr 7:S1535-6108(25)00126-6. doi: 10.1016/j.ccell.2025.03.025. Online ahead of print.

ABSTRACT

Brain metastases (BrMs) remain a major clinical and therapeutic challenge in patients with metastatic cancers. However, advances in our understanding of BrM have been hampered by the constrained sample size and resolution of BrM profiling studies. Here, we perform integrative single-cell RNA sequencing analysis on 108 BrM samples and 111 primary tumor (PTs) samples to investigate the characteristics and remodeling of cell states and composition across cancer lineages and subsets. Recurring and enriched features of malignant cells are increased chromosomal instability, marked proliferative and angiogenic hallmarks, and adoption of a neural-like BrM-associated metaprogram. Immunosuppressive myeloid and stromal subsets dominate the BrM tumor microenvironment, which are associated with poor prognosis and resistance to immunotherapy. Furthermore, five distinct BrM ecotypes are identified, correlating with specific histopathological patterns and clinical characteristics. This work defines hallmarks of BrM biology across cancer types and suggests that shared dependencies may exist, which may be exploited clinically.

PMID:40215980 | DOI:10.1016/j.ccell.2025.03.025

Cancer Cell. 2025 Apr 7:S1535-6108(25)00132-1. doi: 10.1016/j.ccell.2025.03.031. Online ahead of print.

ABSTRACT

Cuproptosis is a recently identified form of copper-dependent cell death. Here, we reveal that radiotherapy (RT) induces cuproptosis in cancer cells, independent of apoptosis and ferroptosis, and depletes lipoylated proteins and iron-sulfur (Fe-S) cluster proteins-both hallmarks of cuproptosis-in patient tumors. Mechanistically, RT elevates mitochondrial copper levels by upregulating copper transporter 1 (CTR1) and depleting mitochondrial glutathione, a copper chelator, thereby triggering cuproptosis. Integrated analyses of RNA sequencing (RNA-seq) from radioresistant esophageal cancer cells and single-cell RNA-seq from esophageal tumors of patients unresponsive to RT link radioresistance to the downregulation of BTB and CNC homology 1 (BACH1). This downregulation de-represses the expression of copper-sequestering metallothionein (MT) 1E/X, thereby mitigating cuproptosis and contributing to radioresistance. Copper ionophore treatment sensitizes radioresistant cancer cells and cell line- and patient-derived xenografts to RT by potentiating cuproptosis. Our findings unveil a link between RT and cuproptosis and inform a therapeutic strategy to overcome tumor radioresistance by targeting cuproptosis.

PMID:40215978 | DOI:10.1016/j.ccell.2025.03.031

Cell Rep Methods. 2025 Apr 7:101022. doi: 10.1016/j.crmeth.2025.101022. Online ahead of print.

ABSTRACT

Understanding the interplay among clinical variables-such as demographics, symptoms, and laboratory results-and their relationships with disease outcomes is critical for advancing diagnostics and understanding mechanisms in complex diseases. Existing methods fail to capture indirect or directional relationships, while existing Bayesian network learning methods are computationally expensive and only infer general associations without focusing on disease outcomes. Here we introduce random walk- and genetic algorithm-based network inference (RAMEN), a method for Bayesian network inference that uses absorbing random walks to prioritize outcome-relevant variables and a genetic algorithm for efficient network refinement. Applied to COVID-19 (Biobanque québécoise de la COVID-19), intensive care unit (ICU) septicemia (MIMIC-III), and COPD (CanCOLD) datasets, RAMEN reconstructs networks linking clinical markers to disease outcomes, such as elevated lactate levels in ICU patients. RAMEN demonstrates advantages in computational efficiency and scalability compared to existing methods. By modeling outcome-specific relationships, RAMEN provides a robust tool for uncovering critical disease mechanisms, advancing diagnostics, and enabling personalized treatment strategies.

PMID:40215965 | DOI:10.1016/j.crmeth.2025.101022