Science Bulletin

Despite of the highly potent antiretroviral therapies, HIV-1 establishes persistent infection and causes chronic inflammation in AIDS patients. Beyond CD4+ T cells, HIV-1 infects myeloid cells, including circulating monocytes and tissue-resident macrophages, and integrates with host genomes to form stable viral reservoirs. To achieve a functional HIV cure, latency-promoting agents (LPAs) have been developed for the "block-and-lock" strategy to reinforce deep HIV-1 latency and permanently silence proviruses. However, most LPAs have been tested mainly in CD4+ T cells, and their efficacy in myeloid cells remains unclear. In this study, we reported that levosimendan (LSM), a drug approved for clinic use to treat heart failures, is able to inhibit HIV lytic infection and reactivation in myeloid cells. LSM blocked viral lytic reactivation in HIV-1 latently infected monocytic cells (TH89GFP, U1) and microglial cells (HC69). LSM also inhibited HIV infection in human induced pluripotent stem cell (iPSC) derived microglia (iMG), primary human resident liver macrophages (Kupffer cells) as well as human monocyte-derived macrophages (MDMs). Furthermore, we demonstrated that overexpression of a predicted drug target of LSM, the conserved serine/threonine kinase RIOK1 (RIO kinase 1), overcomes LSMs anti-HIV effect. Overall, our studies concluded that LSM is a promising LPA to inhibit HIV-1 infection in myeloid cells in the RIOK1-dependent manner.

Astrocytes play a key role in neuronal homeostasis and in various neural disorders. The generation of astrocytes from neural progenitor cells (NPCs) and its functions are under a complex control of several signaling networks and transcription factors. In this study, we demonstrate that the transcription factor, GLIS similar 3 (GLIS3), which has been implicated in several neurodegenerative diseases, is highly expressed in astrocytes, and is required for the efficient differentiation of human NPCs into astrocytes. Loss of GLIS3 function greatly impairs astrocytes differentiation, resulting in reduced expression of astrocyte markers, whereas expression of exogenous GLIS3 restores the induction of astrocyte specific genes indicating a critical role for GLIS3 in astrocyte differentiation. Integrated transcriptomic and cistromic analyses revealed that GLIS3 directly regulates the transcription of several astrocyte-associated genes, including GFAP, SLC1A2, NFIA, and ATF3, in coordination with lineage-determining factors, such as STAT3, NFIA, and SOX9. We hypothesize that GLIS3 dysfunction disrupts this transcriptional network thereby contributing to astrocyte-associated neurological disorders. Identification of GLIS3 as a key regulator of astrocyte differentiation and gene expression will advance our understanding of its role in neurodegenerative diseases and may provide a new therapeutic target.

African swine fever (ASF) is a highly pathogenic disease caused by the African swine fever virus (ASFV) infection, which can affect pigs of all ages and breeds, posing significant threat to the global pig farming industry. The ASFV p30 protein is an early-expressed viral structural protein; however, its function is not fully understood. In this study, the interaction of viral p30 with host TRIM21 was identified. The ectopic TRIM21 inhibited ASFV replication, while knockdown or knockout of TRIM21 promoted ASFV replication. Further, p30 was found to interact with RIG-I-like receptor (RLR) signaling adaptor MAVS, and during ASFV infection, p30-TRIM21-MAVS interacted with each other. Mechanistically, TRIM21 activated the K27 polyubiquitination of MAVS to induce IRF3 mediated type I interferon (IFN) production, whereas p30 counteracted TRIM21 activated MAVS K27 polyubiquitination to evade RLR signaling mediated antiviral IFN induction. In summary, our study revealed a novel function of ASFV p30, and provided new insights into the immune evasion of ASFV.

The impact of cellular activities on large-scale brain dynamics is thought to determine brain functioning and disease, yet the causal relationships of neural mechanisms across scales remain unclear. Recently, the cerebellum has been reported to affect whole-brain dynamics during sensorimotor integration. To disclose the underlying mechanisms, we have developed a multiscale digital brain co-simulator, in which a spiking neural network of the olivo-cerebellar microcircuit is embedded in a mouse virtual brain and wired with other nodes using an atlas-based long-range connectome. Parameters and bi-directional interfaces between the spiking olivo-cerebellar network and other rate-coded modules were tuned to match experimental data of primary sensory and motor cortex (M1 and S1) power spectral densities and neuronal spiking rates. Then, the role of the cerebellar circuitry on sensorimotor integration was analyzed by lesioning critical circuit connections in silico. Simulations showed that spike processing within the cerebellar circuit is key to explaining the gamma-band coherence between M1 and S1 during sensorimotor integration. These results provide a mechanistic explanation of how the cerebellum promotes the formation of sensorimotor contingencies in relevant cortical modules as the basis of its critical role in sensorimotor prediction. On a broader perspective, this modelling approach opens new perspectives for the multiscale investigation of brain physiological and pathological states in relation to specific cellular and microcircuit properties.

Neonates requiring intensive care are at increased risk for long-term neuropsychiatric disorders. However, clinical adoption of risk prediction models remains limited when their performance lacks adequate interpretability for informed clinical decision-making. Here, we investigated whether longitudinal neonatal electronic health record (EHR) data from the first 90 days of life can support clinically meaningful interpretation of long-term risk signals for major neuropsychiatric diagnoses by age seven. In a retrospective register-based cohort of 17,655 at-risk children from an academic medical center, of whom 8.0\% (1,420) received a major neuropsychiatric diagnosis during follow-up, we applied a time-aware transformer model (Self-supervised Transformer for Time-Series; STraTS) and thoroughly evaluated its predictions using three complementary interpretability approaches: perturbation-based variable importance, value-dependent effect analysis, and leave-one-out (LOO) feature attribution. STraTS achieved the highest area under the precision--recall curve (AUPRC 0.171 {+/-} 0.022), compared with Random Forest (0.166 {+/-} 0.008), logistic regression (0.151 {+/-} 0.007), and XGBoost (0.128 {+/-} 0.010). Across interpretability methods, five predictors were consistently identified: birth weight, gender, Apgar score at 1 minute, umbilical serum thyroid stimulating hormone (uS-TSH), and treatment time in hospital. Indicators of early clinical severity, including chromosomal abnormalities and neonatal cerebral-status disturbances, showed the largest risk-increasing effects. Furthermore, the model's learned vector representations of subject-specific EHR sequences formed clinically coherent latent embeddings that reflect population heterogeneity along established perinatal risk dimensions. These findings demonstrate that combining multiple complementary interpretability methods yields stable, clinically plausible risk signals while revealing limitations that would remain undetected by any single approach, highlighting the importance of careful interpretability analysis of deep learning-based risk predictions.

Glioma is the most common primary malignant tumor of the brain, and accumulating evidence indicates that neuronal activity plays a pivotal role in tumor progression. In this study, neuronal activity is modulated in vitro using potassium chloride (KCl)-induced depolarization and midazolam (MDZ)-mediated suppression. MDZ is a neuronal activity modulation medication, commonly used for sedation, anxiolysis, and amnesia in clinics. After treatment, conditioned media derived from these neuronal cultures are subsequently co-cultured with glioma cells. EdU incorporation assays demonstrate that MDZ significantly inhibits glioma cell proliferation in vitro. Furthermore, an orthotopic xenograft glioma model is established to assess the anti-tumor efficacy of MDZ in vivo, as evaluated by tumor volume and Ki-67 immunostaining. Mechanistically, insulin-like growth factor 1 (IGF1) is identified as the neuronal-activity-regulated factor that promotes glioma growth through activation of the PI3K/AKT signaling pathway. Moreover, transcriptomic profiling of brain tissues reveals that MDZ attenuates neuronal activity and downregulates neuron-derived growth factors in both glioma and non-tumor regions, thereby exerting anti-tumor effects in vivo. Collectively, these findings demonstrate that MDZ suppresses glioma progression by suppressing neuronal activity and inhibiting neuron-derived trophic factors, providing new insights into the development of therapeutic strategies for glioma.

ObjectiveAiming at the core problems prevalent in biomedical research, including the "translational distance", the difficulty in aligning cross-scale studies, and the lack of direct validation of single-cell systems biology models in human samples, this study aims to verify whether the results of transcriptome-wide Mendelian randomization (TWMR) based on large-scale populations are consistent with the causal inference results of deep learning combined with double machine learning (DML) using single-cell transcriptome data from human samples, to clarify whether statistical biology and systems biology can converge to the same biological truth, and provide methodological support for mechanism dissection and precision medicine research of complex diseases such as rheumatoid arthritis (RA). MethodsThis study integrated multi-omics data to conduct a two-stage causal inference and cross-scale validation analysis. In the first stage, based on the summary statistics of RA genome-wide association study (GWAS) from 456,348 individuals of European ancestry in the UK Biobank (UKB), and cis-expression quantitative trait locus (cis-eQTL) data from 31,684 individuals in the eQTLGen Consortium, a two-sample Mendelian randomization approach was adopted. Transcriptome-wide causal effect analysis was performed using the inverse-variance weighted (IVW) method, MR Egger regression, and weighted median method, and gene-level causal effect values were obtained after strict quality control and multiple testing correction. In the second stage, based on single-cell RNA sequencing (scRNA-seq) data from RA patients and healthy controls (RA group: 11 samples, 211,867 cells; Healthy control group: 38 samples, 456,631 cells), after preprocessing via the Seurat pipeline, batch effect correction, and cell type annotation, a hierarchical deep neural network was constructed to complete feature compression of high-dimensional expression data, and the DML framework was used to estimate the causal effects of genes on RA disease status. Finally, Pearson correlation analysis was performed to conduct cell type-specific cross-scale validation of gene-level causal effect values obtained by the two methods, and the validated model was used to quantify the causal effects of 16 RA-related pathways from the Reactome database. ResultsThis study confirmed that the gene causal effect values obtained from large-scale population TWMR analysis were significantly correlated with those calculated by the deep learning combined with DML model based on single-cell transcriptome data. Among them, the correlation was extremely significant (p<0.001) in core naive B cells (r=0.202, p=3.2e-05, n=414) and core naive CD4 T cells (r=0.102, p=0.037, n=412). The validated DML model successfully quantified the cell type-specific causal effect values of 16 RA-related signaling pathways. ConclusionStatistical biology and systems biology can converge to the same biological truth. The cross-scale consistency between the two can significantly shorten the "translational distance" in biomedical research, and realizes the direct validation of the single-cell systems biology causal model of human samples based on large-scale population genetic data, getting rid of the excessive dependence on animal/cell experimental models in traditional research. This research paradigm not only provides a new path for mechanism dissection and therapeutic target screening of complex diseases such as RA, but also provides a feasible solution for rare disease research to break through the limitation of GWAS sample size, and lays an important theoretical and methodological foundation for constructing standardized systems biology models of human complex diseases and promoting the development of precision medicine.

Expert pharmaceutical chemists interpret molecular structures through a sophisticated cognitive hierarchy, transitioning from local functional moieties to spatial pharmacophores and, ultimately, to macroscopic pharmacological and physicochemical profiles. However, conventional Graph Neural Networks frequently overlook this high-level chemical intuition by treating molecules as single-scale atomic topology. To bridge this gap between human expertise and computational inference, we propose PyrMol, a knowledge-structured pyramid representation learning framework. By constructing heterogeneous hierarchical graphs, PyrMol orchestrates information flow across atomic, subgraph, and molecular levels. Crucially, the subgraph level systematically integrates three complementary expert views comprising functional groups, pharmacophores, and retrosynthetic fragments. To harmonize these explicit domain priors with implicit computational semantics, we introduce an adaptive Multi-source Knowledge Enhancement and Fusion module that dynamically balances their complementarity and redundancy. A Hierarchical Contrastive Learning strategy further ensures cross-scale semantic consistency. Empirical evaluations across ten benchmark datasets demonstrate that PyrMol outperforms 12 state-of-the-art baselines. Furthermore, its "plug-and-play" versatility provides a framework-agnostic performance boost for existing GNN architectures. PyrMol thus establishes a principled data-knowledge dual-driven paradigm for AI-aided Drug Discovery, effectively leveraging domain knowledge to catalyze advances in molecular property prediction.

Providing early diagnosis and personalized treatment for psychiatric disorders like schizophrenia remains challenging, due to important interpersonal differences and still elusive neuronal mechanisms. Whole-brain network models show promising results with clinical relevance for individualized treatment recommendations in neurological disorders. However, their applicability to psychiatry is still limited as models fail to account for inter-individual differences in the correlation structure of brain dynamics. What physiological mechanisms should models incorporate to better account for individual profiles of brain dynamics in schizophrenia patients and healthy controls? Our study compares various metrics of white matter structure and microstructure to inform connection weights between regions. To do so, we inferred regional parameters of whole-brain mean-field models with The Virtual Brain simulator to account for empirical functional connectivity from resting-state functional magnetic resonance imaging of schizophrenia patients and healthy controls. We found that using global fractional anisotropy or apparent diffusion coefficient of white matter fibers to inform the weights in neural mass models can drastically improve model performance. The data-model correlations of simulated and empirical data were significantly improved (from 0.2 to 0.7) over using the number or density of fibers as in many state-of-the-art methods. This approach allows us to uncover personalized maps of excitation-inhibition imbalance, hypothesized to underlie symptoms in schizophrenia. These maps prove meaningful in that they can predict diagnosis better than model-independent neuroimaging benchmarks. Our findings highlight the importance of white matter microstructure in whole-brain modeling. The novel white-matter-informed models reveal mechanisms that can cause altered brain dynamics in schizophrenia and could inform treatment in personalized psychiatry.

AI-based molecular dynamics simulation brings ab initio calculations to biomolecules in an efficient way, in which the machine learning force field (MLFF) locates at the central position by accurately predicting the molecular energies and forces. Most existing MLFFs assume localized interatomic interactions, limiting their ability to accurately model non-local interactions, which are crucial in biomolecular dynamics. In this study, we introduce ViSNet-PIMA, which efficiently learns non-local interactions by physics-informed multipole aggregator (PIMA) and accurately encodes molecular geometric information. ViSNet-PIMA outperforms all state-of-the-art MLFFs for energy and force predictions of different kinds of biomolecules and various conformations on MD22 and AIMD-Chig datasets, while adapting the PIMA blocks into other MLFFs further achieves 55.1% performance gains, demonstrating the superiority of ViSNet-PIMA and the universality of the model design. Furthermore, we propose AI2BMD-PIMA to incorporate ViSNet-PIMA into AI2BMD simulation program by introducing "Transfer Learning-Pretraining-Finetuning" scheme and replacing molecular mechanics-based non-local calculations among protein fragments with ViSNet-PIMA, which reduces AI2BMDs energy and force calculation errors by more than 50% for different protein conformations and protein folding and unfolding processes. ViSNet-PIMA advances ab initio calculation for the entire biomolecules, amplifying the application values of AI-based molecular dynamics simulations and property calculations in biochemical research.

Liquid-liquid phase separation (LLPS) underlies the formation of biomolecular liquid condensates (also referred to membraneless organelles, MLOs), which are essential for spatially organizing various biochemical processes within cells. Proteins that play a key role in driving condensates formation are termed phase-separating proteins (PSPs). Given experimental identification of PSPs remains labor-intensive and time-consuming, multiple computational tools have been developed based on empirical features or deep learning. In this study, we propose SSPSPredictor, a novel multimodal predictive model for PSPs with folded or intrinsically disordered structures, leveraging the fusion of sequence information from a protein language model ESM-2 and structural insights from a graph neural network GVP. Compared with existing tools, SSPSPredictor achieves balanced performance in identifying endogenous PSPs, predicting relative LLPS propensities, and recognizing key regions that drive LLPS. Moreover, SSPSPredictor exhibits good interpretability in identifying driving regions along protein sequences, although no relevant supervision was provided during training. Further predictive analysis of the human proteome using SSPSPredictor reveals that the proportion of intrinsically disordered proteins (IDPs) undergoing LLPS is significantly higher than that of folded proteins. In addition, pathogenic variants, especially those located in disordered regions, exhibit higher LLPS propensity than other mutations, uncovering a link between LLPS and diseases at the amino acid level.

Neurofibromin 1 (NF1) is a critical negative regulator of the RAS-RAF-ERK pathway, mutations in which have been clinically implicated in various neurodevelopmental disorders. However, the lack of a high-resolution spatiotemporal map has obscured the understanding of why specific cell populations and developmental processes are uniquely vulnerable to NF1 loss. In this study, we present a comprehensive atlas of NF1 expression in the developing mouse brain. Using in situ hybridization and immunohistochemistry, we characterized NF1 distribution from early embryonic stages through postnatal maturation. We further integrated these findings with single-nuclei RNA-sequencing (snRNA-seq) datasets from adult mouse brain to achieve higher resolution. Our results reveal a previously undocumented graded expression pattern of NF1 across various brain regions and lineages. This comprehensive study will not only help in understanding the fundamental role of NF1 during brain development but will also be pivotal in providing a framework to study NF1-associated brain disorders.

Brain-body interactions (BBIs) are fundamental to cognition and mental health, but their continuous multimodal dynamics remain difficult to extract. Previous approaches have been largely observational, and few frameworks enable these interacting processes to be modeled within an integrated generative system. Here, we applied a Predictive-Coding-Inspired Variational RNN (PV-RNN) to simultaneous EEG, ECG, and respiration recordings obtained from 33 participants during exteroceptive and interoceptive attention. The model learned a temporal hierarchy spanning modality-specific dynamics, multimodal associative integration, and sequence-level global states, and accurately reconstructed unseen physiological sequences. Specifically, the intermediate associative layer successfully captured the core complexities of BBI by extracting multiscale, nonlinear, and bidirectional coupling dynamics with variable temporal lags. Furthermore, the estimated precision (inverse variance) of latent variables representing BBI dynamics within this multimodal associative layer increased significantly during interoceptive attention. The magnitude of this condition-dependent precision enhancement correlated positively with subjective adaptive body controllability and negatively with psychiatric vulnerabilities, including rumination and trait anxiety. These findings identify a latent physiological signature of interoceptive attention and establish hierarchical generative modeling as an interpretable framework for extracting continuous BBI dynamics and linking multimodal physiology to cognitive and clinical characteristics.

Enhancer-regulating epigenetic modifiers play critical roles in normal physiological processes and human pathogenesis. The major enhancer regulator paralogs MLL3 and MLL4 (MLL3/4) belong to the lysine methyltransferase 2 (KMT2) family, which catalyzes the methylation of lysine 4 on histone H3 (H3K4me). MLL3/4 are required for enhancer activation and are essential for mammalian development and stem cell differentiation. Recent studies have linked MLL3/4 with different metabolic pathways in the context of stem cell self-renewal and cancer cell growth; however, the underlying mechanisms remain elusive. Here, we utilize Seahorse extracellular flux analysis, stable isotope tracing, stem cell biology techniques, and transcriptomic analysis to investigate the functional relationship of MLL3/4, cellular respiration, and stem cell differentiation. Our results indicate that the loss of MLL3/4 impairs glycolytic activity and mitochondrial respiration in murine embryonic stem cells by downregulating the rate-limiting glycolytic enzyme Hexokinase 2 (HK2) and impairing the function of the Alpha-ketoglutarate dehydrogenase (OGDH) complex. Furthermore, simultaneously overexpression of HK2 and OGDH rescues defects in both cellular respiration and differentiation caused by MLL3/4 loss. Taken together, our study reveals a novel mechanism by which epigenetic machineries such as MLL3/4 govern the differentiation of pluripotent stem cells and facilitates the understanding of disease pathogenesis driven by enhancer malfunction.

Recoverin is a key calcium sensor that controls the desensitization of the visual rhodopsin by GRK1. Previous studies have traditionally been conducted on bovine protein (bRec), while data on human ortholog (hRec) remain scarce. Here, we combine X-ray crystallography, X-ray absorption spectroscopy (XANES), quantum mechanical calculations, molecular dynamics, and functional assays to provide an integrated characterization of hRec. The 2Ca2+-bound hRec structure was solved at 1.60 [A], showing that, unlike bRec, hRec interacts with ROS membranes at physiologically relevant submicromolar Ca2+ levels, due to a species-specific charge distribution that might influence membrane interactions. Both recoverins form a set of Ca2+/Zn2+-bound conformers with improved functional performance. X-ray crystallography (1.85 [A]) and XANES revealed a specific tetrahedral Zn2+ site in 1Ca2+-bound hRec, the first such site reported in the NCS family. In 1Ca2+-bound hRec, zinc promotes the formation of active state, whereas in 2Ca2+-state of bRec, it significantly enhances GRK1 binding, as the latter can complement the Zn2+ coordination. These data refine our understanding of recoverin function in humans and highlight its role as a key link between calcium and zinc signaling in mammalian photoreceptors under normal and pathological conditions.

Highly pathogenic avian influenza viruses (HPAIVs) continue to cause substantial disease in birds and mammals, with repeated H5N1 spillovers highlighting the need for broadly protective antiviral strategies. Here we develop a programmable RNA-targeting antiviral platform based on RfxCas13d and evaluate its activity in avian cells. Screening of five Cas13 orthologs in chicken DF1 fibroblasts revealed RfxCas13d as the most potent and well tolerated effector. Virus-specific CRISPR RNAs (crRNAs) targeting conserved regions of positive- and negative-sense influenza RNA were tested against A/WSN/033[H1N1] and multiple HPAIV isolates, including a member of clade 2.3.4.4b H5N1. Targeting positive-sense RNA conferred superior influenza inhibitory activity and further enhanced by multiplexed crRNA expression. These findings establish RfxCas13d as a versatile RNA-guided antiviral platform and provide a route for broad-spectrum influenza control through conserved RNA targeting.

BackgroundMonoamine transporters (MATs), including the dopamine, norepinephrine, and serotonin transporters (DAT, NET, SERT), are essential regulators of synaptic neurotransmission that rely on complex oligomeric bio-assemblies for proper function. The regulatory influence of naturally occurring, alternatively spliced truncated isoforms on bio-assembly dynamics remains profoundly underexplored. MethodsTo decode these interactions at the atomic level, we deployed a multiscale computational framework. We integrated genomics-guided multimer predictions with massive-scale, near-million-atom molecular dynamics (MD) simulations within explicit lipid bilayers. The thermodynamic stability of these heteromeric complexes was quantified using membrane-adapted MM/PBSA calculations, which were subsequently correlated with dynamics-aware evolutionary profiling to map co-evolutionary interaction hotspots. ResultsOur analyses reveal that the NET-derived truncated isoform A0A804HLI4 acts as a pan-family potential inhibitor. It forms stable, exergonic heterodimers with canonical NET and DAT, thermodynamically outcompeting native homodimerization. In full tetrameric simulations, the integration of a single isoform precipitates a macro-structural disruptions of the SERT complex. The variant anchors at non-native interfaces, locking the assembly into an asymmetric, non-native state. Residue-level thermodynamic decomposition and evolutionary mapping isolated conserved structural elements (most notably Gln236) that dictate this high-affinity cross-reactivity across the SLC6 family. ConclusionsTruncated MAT isoforms execute a dynamic mechanism of inhibitory effects and may systematically downregulate synaptic reuptake capacity by sequestering functional monomers. These findings establish a thermodynamically grounded, high-resolution model of isoform-induced bio-assembly disruptions. Crucially, they expose these non-canonical, isoform-driven interfaces as conserved and highly druggable targets, offering a distinct pharmacological paradigm for precision interventions in neuropsychiatric and neurodegenerative pathologies.

Alzheimers disease (AD), the leading cause of dementia, affects over 33 million people worldwide, with pathogenesis tied to amyloid-{beta} (A{beta}) accumulation. Although anti-A{beta} monoclonal antibodies have shown clinical benefits, they often cause side effects including amyloid-related imaging abnormalities and brain microhemorrhage, especially in APOE E4 allele carriers. Here we used PHP.eB serotype adeno-associated virus (AAV), a vector with enhanced central nervous system (CNS) tropism, to deliver an A{beta} antibody expression vector (AAV-LEC) into the CNS of APP/PS1 and 5xFAD mice intravenously. The AAV-LEC-mediated expression of anti-A{beta} antibodies in the CNS significantly reduced the number and size of A{beta} plaques at various stages in both APP/PS1 and 5xFAD mice, alongside improved spatial learning and memory. It also reversed abnormal glial activation with reduced disease-associated microglia and astrocytes, and restored oligodendrocyte differentiation and myelin formation. No brain microhemorrhage or liver damage was detected following the AAV-antibody treatment. Thus, this AAV-mediated strategy offers a promising, convenient and safe AD therapeutic approach in the future.

Cardiac arrhythmias are abnormal heart rhythms characterized by disordered electrical dynamics that impair cardiac function and pose a major global burden of morbidity and mortality. Early and accurate prediction of arrhythmic anomalies from physiological time series is crucial for effective intervention, yet remains challenging due to the nonlinear, nonstationary, and individualized nature of cardiac dynamics. Despite significant advances in machine learning-based arrhythmia detection, most existing methods operate as static classifiers on electrocardiographic signals and lack online prediction, patient-specific adaptation, and mechanistic interpretability. From a dynamical-systems perspective, arrhythmias represent qualitative regime transitions, often preceded by subtle, temporally extended deviations that are difficult to detect in real time. Here we introduce CASCADE (Chaotic Attractor Sensitivity for Cardiac Anomaly Detection), an online and personalized anomaly forecasting framework built on a special type of reservoir computing called Dynamical Systems Machine Learning (DynML). DynML employs ensembles of continuous-time nonlinear dynamical systems as chaotic reservoirs to reconstruct and forecast short-term cardiac dynamics on a beat-to-beat basis, training only a linear readout. This design enables efficient online adaptation without retraining the underlying dynamical model. Rather than relying on static beat-level classification, CASCADE identifies arrhythmic events as failures of short-term predictability, manifested as statistically significant deviations between predicted and observed dynamics relative to subject-specific baselines. Detection performance is governed by the intrinsic dynamical complexity of the reservoir, quantified by topological entropy. Reservoirs operating near critical entropy regimes optimally amplify subtle, temporally extended irregularities in heartbeat dynamics, rendering incipient arrhythmic signatures linearly separable at the readout level. Topological entropy thus serves both as a predictor of model performance and a principled control parameter for reservoir design. When evaluated on the MIT-BIH Arrhythmia dataset, CASCADE achieved consistently high F1 scores, precision, recall, and overall accuracy across diverse patient populations, demonstrating strong generalizability across clinical and real-world settings. By integrating chaotic reservoir computing, entropy-guided tuning, and online personalized forecasting, CASCADE reframes arrhythmia detection as a problem of dynamical regime transition rather than static classification. This perspective provides a scalable, interpretable, and computationally efficient framework for real-time cardiac monitoring and early-warning clinical decision support.

Accurate quality assessment of predicted protein-protein complex structures remains a major challenge. Existing graph-based quality assessment methods often treat the entire complex as a homogeneous graph, which obscures the physical distinction between intra-chain folding stability and inter-chain binding specificity. In this study, we introduce TriGraphQA, a novel triple graph learning framework designed for model quality assessment of protein complexes. TriGraphQA explicitly decouples monomeric and interfacial representations by constructing three geometric views: two residue-node graphs capturing the local folding environments of individual chains, and a dedicated contact-node graph representing the binding interface. Crucially, we propose an interface context aggregation module to project context-rich embeddings from the monomers onto the interface, effectively fusing multi-scale structural features. We conducted comprehensive tests on several challenging benchmark datasets, including Dimer50, DBM55-AF2, and HAF2. The results show that TriGraphQA significantly outperforms state-of-the-art single-model methods. TriGraphQA consistently achieves the highest global scoring correlations and lower top-ranking losses. Consequently, TriGraphQA provides a powerful evaluation tool for protein-protein docking, facilitating the reliable identification of near-native assemblies in large-scale structural modeling and molecular recognition studies.