Biology subjects

Liu, D.

Publications and source records attributed to Liu, D..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

Early Tracheal and Salivary miRNAs in Extremely Preterm Infants Predict BPD-related Pulmonary Hypertension

Pulmonary hypertension (BPD-PH) associated with bronchopulmonary dysplasia (BPD) in preterm infants associates with high morbidity and mortality within the first two years of life. In a previous unbiased study, we identified a panel miRNAs in tracheal aspirates (TA) that were differentially expressed in extremely low gestational age newborns (ELGANs) with BPD-PH compared to those with BPD but no PH. To explore the predictive potential of these miRNAs, we studied TA exosomes from 7 days old ELGANs and analysed a curated panel of 16 miRNAs through logistic regression and calculated the predictive AUROC to diagnose BPD-PH at 36 weeks PMA. AUROC of TA miRNAs was 0.76 with sensitivity and specificity of 53% and 93%, respectively. Adding sex and gestational age to the variables improved the AUROC to 0.78 with sensitivity and specificity of 61 and 87% respectively. Due to challenges of obtaining TA in non-invasively ventilated infants, we collected saliva samples from ELGANs at 7 days of age and compared the log expression of these 16 miRNAs in both biofluids and found significant correlation in their expression (pearson r=0.92, p<0.001). We calculated the predictive AUROC of the same miRNAs to diagnose BPD-PH at 36 weeks PMA. AUROC of these miRNAs in saliva was = 0.85 with sensitivity and specificity of 82% and 72%, respectively; addition of biological sex and gestational age improved AUROC to 0.86 with sensitivity and specificity of 79% and 76% respectively. Leave-one-sample-out sensitivity analysis demonstrated stable training performance with reduced performance in testing samples, supporting the need for validation in larger independent cohorts. In conclusion, early salivary miRNAs have great potential for risk stratification of ELGANs to develop BPD-PH, while also providing the opportunity to identify target molecules and mechanisms that modulate molecular function.

bioinformatics↗

bioRxiv2026

Malaria parasites coordinate host cell remodeling through the AP2-HCR DNA-binding protein

Parasite-induced host cell remodeling is essential for Plasmodium falciparum survival during the intraerythrocytic developmental cycle (IDC), yet the transcriptional mechanisms coordinating this process remain poorly understood. Here, we characterize an essential ApiAP2 transcription factor, PfAP2-HCR (host cell remodeling), during both the asexual and sexual blood stages of P. falciparum. Conditional truncation of pfap2-hcr demonstrated that PfAP2-HCR is essential for ring stage development and parasite maturation. Loss of PfAP2-HCR disrupted the expression of host cell remodeling genes, resulting in the absence of Maurers clefts and collapse of the parasite protein export machinery. Integrated RNA-seq and ChIP-seq analyses revealed that PfAP2-HCR directly or indirectly regulates approximately 30% of exported proteins, including 27 essential exported factors, the parasitophorous vacuole membrane (PVM) resident protein EXP1, and the core PTEX component EXP2. We further show that PfAP2-HCR is also indispensable for gametocytogenesis, as conditional truncation results in either complete ablation or the formation of severely deformed gametocytes. Together, these findings establish PfAP2-HCR as a central transcriptional regulator that orchestrates the expression of host cell remodeling factors required to transform a terminally differentiated erythrocyte into a permissive environment for parasite survival, development, and transmission.

microbiology↗

bioRxiv2026

Establishing a Retron-Based Cytosine Base Editor for Targeted Hypermutation in Escherichia coli

Current cytosine base editors (CBEs) are limited to unidirectional C to T conversions, restricting their applications. Retrons, bacterial genetic elements, encode a reverse transcriptase that generates multicopy single-stranded DNA (msDNA) by reverse transcribing specific non-coding RNA (ncRNA). This msDNA mimics Okazaki fragments during DNA replication, making retrons promising for gene editing. Here, we developed a retron-based cytosine base editor (RCBE) by fusing cytosine deaminase with reverse transcriptase (RT-CDA) within the retron system. RCBE first transcribes ncRNA, allowing RT-CDA to deaminate cytosine on the ncRNA. The modified ncRNA is then reverse transcribed into msDNA, where RT-CDA induces further cytosine deamination. This mutant msDNA introduces specific mutations into target gene sequences, enabling both C to T and G to A conversions. Using RCBE, we demonstrated accelerated molecular evolution of the rpoB gene in Escherichia coli. High-throughput sequencing confirmed that RCBE achieves a mutation rate of up to 0.2% in regions with high GC content. Our findings establish RCBE as a versatile tool, particularly suitable for directed evolution in GC-rich regions, with broad potential applications across various bacterial and eukaryotic hosts.

Synthetic Biology↗

bioRxiv2026

Human Haematopoietic Stem Cells released into circulation following mobilization express multiple G0-associated quiescence markers

Haematopoietic stem cells (HSC), while usually quiescent, can rapidly divide following specific stimuli (mobilization). These HSC can seed additional niches, allowing for the swift generation of essential blood cells. However, studies in mice and humans have clearly demonstrated that cycling bone marrow (BM) HSC (cells in the G1/S/G2/M phases) engraft and reconstitute the haematopoietic system poorly compared with HSC in the G0 phase1. This raises the question why mobilized HSC, immediately following 3 or more cell divisions2, efficiently reconstitute the haematopoietic system. We studied this phenomenon in human HSC using scRNAseq analysis. We found that mobilized HSC rapidly start transcribing genes associated with quiescence, specific for the G0 phase of the cell cycle. We hypothesize that this rapid switch from actively dividing to quiescent cells combined with our extensive RNA expression data will allow us to better define pathways involved in this process.

Cell Biology↗

bioRxiv2026

Microstructural and Biomechanical Determinants of Biological Aging

The pulmonary artery undergoes measurable structural and mechanical deterioration with age, but whether these changes can be integrated into a quantitative normative aging prediction model has not been demonstrated. Using two-photon imaging and paired vascular mechanical measurements from C57BL6 mice spanning 6 to 24 months, we developed a multimodal support vector regression (SVR) model integrating collagen fiber orientation, straightness, and hemodynamic mechanical parameters to predict normative age. Fiber orientation was encoded via the von Mises probability density function referenced to the circumferential and axial vessel wall axes providing a principled circular-variable encoding of both mean direction and concentration. The microstructure-only model achieved leave-one-out (LOO) R{superscript 2} = 0.596, Mean Absolute Error (MAE) = 3.43 months. Adding vascular mechanical parameters (PWV) raised a combined LOO R{superscript 2} to 0.834 (MAE = 2.26 months), a 40.1% improvement. Because pulmonary vascular and parenchymal aging are mechanistically coupled, lung mechanics were included as a complementary readout to assess whether airway mechanics contribute independent predictive signal beyond vascular microstructure alone. A sex dimorphism was observed, where females drove the majority of the collagen-based predictive signal (female-only R{superscript 2} = 0.960 vs. male-only R{superscript 2} = 0.658). These results establish a multimodal framework for vascular biological age quantification that integrates structural and mechanical aging signatures.

bioengineering↗

bioRxiv2026

A dish-to-biobank framework links β-cell nutrient-stress programs to genetic and dietary risk for Type 2 Diabetes

Type 2 diabetes (T2D) arises from genetic susceptibility and chronic metabolic stress, but whether these converge on shared molecular programs in human populations remains unclear. Here, we develop a dish-to-biobank framework linking controlled {beta}-cell perturbation to population-scale disease genetics through the circulating plasma proteome, and apply it to T2D. scRNA-seq of human stem cell-derived islets under factorial glucose and palmitate exposure identifies their combination (glucolipotoxicity) as the condition eliciting the strongest SC-{beta} cell transcriptional response, with glucolipotoxicity-upregulated genes uniquely enriched for T2D heritability, monogenic diabetes genes, and rare-variant burden signals. CRISPR knockout of {beta}-cell identity regulators PAX6 and PDX1 aligns with this program, establishing convergence of environmental and genetic perturbations on a shared disease-relevant state. We then used the plasma proteome as an accessible population-scale readout of these experimentally defined {beta}-cell stress programs, scoring 45,956 UK Biobank White British participants. We define heritable stress signatures that associate with refined carbohydrate and saturated fat intake, and undergo trans-tissue genetic regulation, with a subset of variants showing diet-dependent effects. Together, these findings establish glucolipotoxicity as a genetically anchored model of {beta}-cell dysfunction and provide a generalizable framework for linking controlled cellular perturbations to human disease genetics at population scale.

genetics↗

bioRxiv2026

Biological Aging of the Cardiopulmonary System

Age-related stiffening of large arteries is a predictor of cardiovascular morbidity and mortality, yet how pulmonary vascular stiffening integrates with right ventricular (RV) and lung functional decline--and how best to quantify "biological" cardiopulmonary aging--remains unclear. Here we map cardiopulmonary aging across the adult murine lifespan by integrating RV, proximal pulmonary artery (PA), and lung biomechanics with single-cell transcriptomics. Using ex vivo biaxial testing of the proximal PA, in vivo echocardiography, and lung mechanics, we find that cardiopulmonary aging is phase-dependent: PA circumferential stiffening and reduced distensibility progress largely linearly with age; whereas, RV remodeling and lung mechanical changes exhibit non-linear trajectories. This is consistent with early intrinsic functional decline of cells and organs followed by later, extrinsic load-dependent structural adaptation. To quantify organ-level biological aging, we apply principal component analysis to PA, RV, and lung feature sets to derive physiology-based aging scores that summarize coordinated variance within and across organs. Anchoring differential gene expression in PA single-cell RNA-seq to these continuous biological aging scores rather than chronological age reveals extensive, cell-type-specific remodeling programs (13,636 genes) that are sparse or non-informative when modeled by chronologic age. Biological aging associates across endothelia, smooth muscle cells, fibroblasts, and perivascular macrophages with increased oxidative phosphorylation signatures alongside suppression of adaptive/regulatory pathways, including impaired endothelial mechanotransduction, reduced smooth muscle Wnt signaling, altered extracellular matrix remodeling programs, and erosion of macrophage innate immune and TGF{beta}/NF-{kappa}B signaling nodes. These findings support a model in which pulmonary arterial stiffening is not merely a marker but an active contributor to cardiopulmonary aging via a biomechanical-metabolic-inflammatory uncoupling that diminishes vasoactive and mechano-adaptive reserve and promotes a positive feedback loop. Together, our work establishes physiology-derived biological aging as a powerful framework for interpreting vascular single-cell aging trajectories and identifies mechanistic pathways to target pulmonary vascular stiffening and preserve cardiopulmonary function with age.

physiology↗

bioRxiv2026

Uncovering Pseudotime-Varying Genetic Causal Effects Along Single-Cell Trajectories for Pulmonary Disease Trait

With the increasing accessibility of single-cell RNA sequencing (scRNA-seq) data, cell-type-specific gene expression can be linked to complex traits through pseudo-bulk method, which considered aggregated gene expression from multiple cells of the same annotated cell type per individual and clearly shows the limitation of ignoring intra-individual cell-to-cell variability. Concurrently, pseudotime trajectory inference has gained popularity for its ability to capture continuous biological processes such as cell differentiation and lineage development, instead of individual discrete stages. It is natural to consider whether genetic effects for complex traits, such as individual level disease status, show a dynamic pattern along the inferred trajectories. In this study, we introduce a novel framework that models gene expression as a function of pseudotime along the inferred trajectories. We mapped expression quantitative trait loci (eQTL) effects in the cis-region as functional parameters, which we called "dynamic eQTLs", showing regulatory effects exerted by genetic variants change continuously along the cellular trajectory. For eQTLs of constant effects across pseudotime we leveraged external bulk-eQTL information to enhance the power. Furthermore, we employed significant, variable dynamic eQTLs as instrumental variables to infer causal relationships between gene expression and complex traits. To address challenges inherent to scRNA-seq data--such as sparsity and high variability--we incorporate an empirical likelihood-based inference method, which is non-parametric and self-normalized. Besides, genes associated with trajectory branchpoints may bring confounding, and we also proposed a causal mediation analysis framework to determine whether a gene plays a causal role for the disease directly and indirectly through driving cell fates. Applying our method to scRNA-seq data from human lung tissue of 114 samples (66 pulmonary fibrosis cases and 48 controls), along with meta-analyzed GWAS summary statistics for IPF from 3 studies, we identified pseudotime-dependent causal effects for IPF from genes implicated in the trajectory AT2 - translational AT2 - AT1, which is crucial in lung tissue repair and regeneration. We also found that 30 genes have a mediated effect through cell fates.

genetics↗

bioRxiv2026

An AI-agent-orchestrated grey-box Transformer framework for sparse pharmacokinetic curve reconstruction and pharmacometric model initialization

Clinical pharmacokinetic (PK) modelling is constrained by sparse sampling, limited general-isability of single-drug models, and labour-intensive workflows, making it difficult to infer complete drug exposure from limited concentration observations. We present the Pharmacokinetic Foundation Model (PKFM), a grey-box Transformer framework pre-trained across 32 drugs that reconstructs concentration-time profiles from sparse concentration observations, dosing events, molecular descriptors, and physiological covariates while preserving output interpretability. In representative oral PK curves, three sparse input points recovered the principal absorption-elimination trajectory, achieving coefficient of determination (R2) = 0.992 for Midazolam oral and R2 = 0.990 for Verapamil oral. Using reconstructed curves in NONMEM (nonlinear mixed-effects modelling) improved covariance stability and individual prediction accuracy. Contrastive-learning embeddings supported Top-10 physiologically based pharmacokinetic (PBPK) candidate retrieval, with 75.6% of observations within the 2-fold range. A pharmacometrics-informed AI Agent (PM Agent) outperformed general-purpose programming tools in stability and pairwise win rate on a standardised modelling benchmark, with each run requiring human pharmaco-metrician confirmation before downstream use. These results support cross-drug pre-trained PK models as an information-completion layer for sparse PK evidence and a structured scaffold for the modelling workflow; clinical or regulatory use requires prospective validation, broader external benchmarking, and independent expert assessment.

pharmacology and toxicology↗

bioRxiv2026

GeoEPred: A Multimodal Structure-Aware Geometric Deep Learning Framework for Gram-Negative Bacterial Secreted Effector Prediction with Sequence Semantics

Accurate prediction of effector proteins secreted by Gram-negative bacteria is important for elucidating bacterial pathogenic mechanisms and developing precise anti-infective strategies. Although existing methods have benefited from the strong sequence feature extraction capacity of pretrained protein language models, reliance on linear sequence information alone often fails to fully capture the three-dimensional conformational signals required for virulence functions. Meanwhile, conventional structure-based methods are limited by the scarcity of experimentally resolved protein structures. To address these challenges, We propose GeoEPred, a multimodal deep learning framework designed for the synergistic modeling of protein sequence and structure to identify Gram-negative bacterial effector proteins. Specifically, the model integrates sequence-contextual embeddings from a pretrained protein language model with three-dimensional structural representations predicted by ESMFold. A feature projection network refines fine-grained sequence signals associated with effector functions, while geometric vector perceptrons characterize inter-residue orientations, distances, and local spatial topology to capture potential structural conformational motifs. To further enable effective cross-modal fusion, we design a cross-modal alignment and feature-tokenized self-attention module. This module enhances consistency between the sequence-semantic and structural-geometric spaces through contrastive learning and models associations between linear functional motifs and spatial conformational patterns at a fine-grained token level. Extensive evaluations on multiple benchmark datasets show that GeoEPred achieves better predictive performance than existing leading models in T3SE, T4SE, and T6SE prediction tasks, while maintaining stable performance in remote homolog recognition scenarios. Moreover, the modular and extensible architecture of GeoEPred demonstrates strong generalization ability and substantial application potential for genome-scale effector protein discovery. Author summarySecreted effector proteins are central virulence factors used by many Gram-negative bacterial pathogens to execute infection strategies. Their functions are governed not only by secretion signals and short linear motifs in the amino acid sequence, but also by three-dimensional folds, local domains, and surface geometric patterns. However, current predictors mainly exploit sequence-contextual features, limiting their ability to model the correspondence between linear sequence signals and spatial conformational motifs, and thereby constraining accuracy and interpretability. Here, we present GeoEPred, a multimodal deep learning framework for secreted effector protein identification. GeoEPred couples sequence-semantic embeddings from a pretrained protein language model with structural representations learned by geometric vector perceptrons. A cross-modal alignment and interaction module uses contrastive learning to improve functional consistency between sequence and structure modalities, while feature-token attention captures fine-grained links between key linear and conformational motifs. Across benchmark datasets covering multiple effector types, GeoEPred outperforms existing state-of-the-art methods and provides interpretable evidence from sequence fragments, structural regions, and cross-modal associations, supporting functional annotation, pathogenic mechanism analysis, and experimental validation.

genomics↗

bioRxiv2026

Genomic, Clinical, and Spatial Predictors of Durable Response to BRAF/MEK Inhibition in BRAF-Mutant Melanoma

BRAF-targeted therapy (BRAFi/MEKi) and immune checkpoint blockade (anti-PD-1/anti-CTLA-4) have transformed the treatment of BRAF-mutant metastatic melanoma. While most patients who respond to targeted therapy eventually progress, a subset derives durable benefit, and biomarkers to identify this subset would inform optimal treatment selection. In this study, we analyzed pre-treatment tumor samples from a clinically annotated cohort of 155 patients with BRAF-mutant metastatic melanoma treated with first-line BRAFi/MEKi and followed for up to five years. We stratified patients into durable responders (PFS [≥] 24 months) and rapid progressors (PFS < 6 months with progression) and found that a global metric of tumor genomic heterogeneity, rather than individual gene alterations, distinguished these groups. Combining genomic heterogeneity with baseline tumor burden (e.g., lactate dehydrogenase (LDH) or radiographic lesion dimensions), we developed a parsimonious model that predicted durable responders with high precision and specificity. Notably, the analogous population of patients treated instead with immunotherapy were not durable responders, suggesting that the selected predictors of durable responders are targeted therapy specific. Spatial profiling of a subset of pre-treatment biopsies (n = 47) demonstrated that high intratumoral, but not peritumoral, CD8+ T-cell infiltration correlated with prolonged survival on BRAF-targeted therapy and served as an independent predictive factor when considered with genomic heterogeneity and features of clinical tumor burden. Together, these findings highlight the distinct baseline intrinsic and extrinsic features underlying durable response to BRAF-targeted therapy and support their potential implication in guiding treatment selection for patients with BRAF-mutant metastatic melanoma. One-Sentence SummaryIntegrated clinical, tumor genomic, and immune microenvironmental features predict durable responses to BRAF-targeted therapy.

cancer biology↗

bioRxiv2026

Thoughts-as-Planning: Latent World Models for Chain-of-Thoughts Optimization via Reinforcement Planning

The success of large language models (LLMs) across diverse NLP tasks has elevated the importance of reasoning chain optimization as a critical step in aligning model behavior with task objectives. Existing reasoning chain tuning methods often rely on black-box heuristics or gradient-free search, which lack interpretability, generalization, and sample efficiency. In this work, we introduce Thoughts-as-Planning, a novel framework that formalizes reasoning chain optimization as a sequential decision-making process over a latent semantic space. We model the LLM as a partially observable environment and learn a latent world model that simulates the effect of reasoning chain edits on downstream outputs. A proximity-preserving embedding space is constructed to encode reasoning chain-response dynamics, enabling planning via gradient descent or reinforcement learning. Our method supports multi-scale abstraction, allowing reasoning chain edits at token, segment, and instruction levels to be integrated into a unified planner. Through extensive experiments on language understanding and generation tasks, we demonstrate that Thoughts-as-Planning outperforms state-of-the-art reasoning chain tuning baselines in efficiency, robustness, and generalization, while offering interpretability through its structured planning trajectory. Our code is available at https://github.com/FastLM/Thoughts-as-Planning.

neuroscience↗

bioRxiv2026

Aldoximes serve as auxin precursors and repress phenylpropanoid metabolism in tomato

Aldoximes are amino acid-derived metabolites that serve as precursors of auxins and modulate phenylpropanoid production in Arabidopsis. However, the enzymes responsible for aldoxime production in Solanaceae remain unknown. Here, we report the identification of aldoxime-producing enzymes in tomato (Solanum lycopersicum) and examine how altered aldoxime production affects auxin production and phenylpropanoid metabolism. Through homology-based analysis, we identified five putative CYP79 homologs in tomato, among which SlCYP79DB32 and SlCYP79DB52 exhibited aldoxime-producing activity toward multiple amino acids, including phenylalanine and tryptophan. SlCYP79DB32 and SlCYP79DB52 converted phenylalanine into phenylacetaldoxime (PAOx), whereas only SlCYP79DB52 converted tryptophan into indole-3-acetaldoxime (IAOx). Stable isotope-labeled feeding experiments revealed that IAOx and PAOx can be converted to the auxins indole-3-acetic acid (IAA) and phenylacetic acid (PAA), respectively. Consistently, tomato plants engineered to overproduce IAOx and PAOx accumulated elevated levels of IAA and PAA. These plants also accumulated lower levels of phenylpropanoids. In Brassicaceae plants such as Arabidopsis and Camelina, aldoxime accumulation represses phenylpropanoid production by promoting degradation of phenylalanine ammonia-lyase (PAL). However, aldoxime accumulation did not reduce PAL activity in tomato, suggesting an alternative mechanism in this species. Transcriptome analysis revealed extensive transcriptional reprogramming in aldoxime-overaccumulating tomato plants, including upregulation of stress- and defense-related genes. Despite the observed reduction in phenylpropanoid levels, transcript levels of most phenylpropanoid biosynthetic genes were not decreased, suggesting possible post-transcriptional regulation of this repression. Together, our findings demonstrate that aldoximes can serve as intermediates in auxin biosynthesis in tomato and reveal that aldoxime-mediated repression of phenylpropanoid metabolism extends beyond Brassicaceae.

plant biology↗

bioRxiv2026

Distinct HPO axis responses and ovarian aging trajectories to chronic unpredictable mild stress in reproductively young versus middle-aged female mice

Psychosocial stressors are key contributors to ovarian functional decline. Chronic unpredictable mild stress (CUMS) is widely used to model stress-induced premature ovarian insufficiency (POI) in mice; however, current animal models do not adequately reflect middle-aged women, who represent a key population exposed to chronic psychosocial stress, nor do they capture the dynamic progression toward POI. Here, female C57BL/6 mice aged 2 or 6 months were subjected to CUMS for 8 or 12 weeks. Estrous cyclicity, endocrine profiles, ovarian histology, and transcriptomic changes in HPO axis-related tissues were systematically analyzed. After 8 weeks of exposure, 2-month-old mice exhibited impaired pituitary responsiveness to estradiol negative feedback, as evidenced by dysregulated FSH secretion, indicating reduced stress tolerance compared with 6-month-old mice. Following 12 weeks of CUMS exposure, both age groups showed significant reductions in ovarian size and follicle numbers across all developmental stages. These findings demonstrate that CUMS induces an age-dependent progression toward POI, with short-term exposure eliciting compensatory phases preceding overt ovarian insufficiency, accompanied by distinct endocrine and reproductive alterations and differential responsiveness of the HPO axis. Transcriptomic analyses revealed age-dependent stress responses: ovaries of 2-month-old mice displayed marked activation of inflammatory and immune-related pathways, whereas 6-month-old mice showed sustained upregulation of protein kinase-related signaling networks. Notably, the 6-month-old CUMS model more closely recapitulates stress-associated reproductive aging in adult women. In briefCUMS has been widely used to establish mouse models of psychosocial stress-induced POI. However, current animal models do not adequately reflect middle-aged women, who represent a key population exposed to chronic psychosocial stress, nor do they capture the dynamic progression toward premature ovarian insufficiency (POI). In this study, we demonstrate that different durations of CUMS exposure induce distinct stages of ovarian dysfunction in both young and middle-aged mice, with short-term exposure driving age-dependent compensatory phases and prolonged exposure leading to overt POI, both accompanied by divergent endocrine and reproductive alterations, alongside age-dependent changes in HPO axis responsiveness to CUMS. Notably, the 6-month-old CUMS model shows greater clinical relevance in recapitulating chronic psychosocial stress and stress-related reproductive aging in adult women.

physiology↗

bioRxiv2026

CellPulse: A Foundation Model of Coordinated Gene Dynamics Simulating Viral Infectious Diseases

Understanding how cells respond to perturbations like viral infections requires models capturing coordinated gene dynamics. However, current gene expression foundation models are predominantly reliant on single-cell data and static gene expression, limiting their applicability in real clinical scenarios. We present CellPulse, a direction-aware foundation model trained on the Virus Stimulated Atlas (VISTA), a newly curated atlas of over 23 million bulk RNA-sequencing differential expression profiles from viral infections. CellPulse models the direction and magnitude of gene expression changes via a structured representation of differential expression and a direction-aware attention mechanism, enabling the learning of coherent regulatory programs. It shows powerful diagnosing capability by accurately classifying 31 distinct virus types across diverse clinical and laboratory samples, solely from host transcriptional signatures. Crucially, without prior knowledge injection, CellPulses interpretability reveals virus-associated host factors that mediate infection. Using a selection of host factors for in silico drug screening yielded numerous compounds with confirmed efficacies in wet-lab assays, while cell-based and animal experiments further verified the causal relationship between host targets and viral infections. Overall, CellPulse represents a generalizable foundation model for deciphering coordinated gene dynamics from bulk transcriptomics, bridging host response modeling with clinical relevance and therapeutic discovery for infectious diseases and beyond.

bioinformatics↗

bioRxiv2026

Expanding P-NET, a multi-purpose biologically informed deep learning framework

We present expanded P-NET, a versatile frame-work for deep learning in computational biology based on P-NET, leveraging biological pathways for interpretable predictions. Our framework achieves competitive performance in genomic & transcriptomic prediction tasks. We demonstrate its stability and interpretability compared to traditional machine learning models. P-NET 2.0 incorporates gene and pathway information, providing valuable insights into complex biological processes. The framework is publicly available, enabling its application to various computational biology tasks.

bioinformatics↗

bioRxiv2026

Highly Accurate Estimation of the Fold Accuracy of Protein Structural Models

BackgroundThe function of a protein is intrinsically linked to its three-dimensional fold, and deep learning has revolutionized the field by enabling high-accuracy structure prediction at an unprecedented scale. Nevertheless, the growing deployment of these predictive pipelines in drug discovery and structural biology reveals a critical bottleneck that lies in the lack of independent and rigorous estimation of model accuracy (EMA) methodologies. ResultsHere we present DeepUMQA-Global, a single-model deep learning framework for estimating accuracy of protein structure models. Our method employs a structure-sequence cross-consistency mechanism to evaluate the bidirectional compatibility between the predicted structure and the input sequence, enabling comprehensive characterization of fold accuracy. DeepUMQA-Global outperforms the self-assessment confidence scores of AlphaFold3, achieving improvements of 57.8% in Pearson correlation and 49.0% in Spearman correlation. With respect to the CASP16 retrospective benchmark, DeepUMQA-Global outperforms all single-model accuracy estimation methods that participated in CASP16 and achieves performance comparable to that of the top consensusObased methods. A lightweight consensus strategy built upon DeepUMQA-Global ranks first among all CASP16 participants, surpassing all other methods, including consensus approaches, and highlighting the strength of our method. Remarkably, DeepUMQA-Global demonstrates a strong ability to discriminate between alternative conformational states of proteins, as evidenced in the CASP unique alternative conformation protein complex target and the CoDNaS benchmark. ConclusionsOur results indicate that DeepUMQA-Global can be extended to broader protein modeling tasks, moving beyond static evaluation to offer a foundation for dynamic conformation EMA, where it accurately discriminates alternative conformational states and demonstrates reliable predictive fidelity in model accuracy estimation.

bioinformatics↗

bioRxiv2026

Bacterial endosymbionts enhance fungal virulence by disrupting the disease-suppressive rhizobiome

The mechanisms by which bacterial endosymbionts (bacteria living within fungal cells) enhance the fitness and virulence of fungal pathogens remain poorly understood. Here, we report that the tomato Fusarium wilt pathogen Fusarium oxysporum f. sp. lycopersici (FOL) hosts Achromobacter sp. endosymbionts that enhance fungal virulence. This virulence potentiation is partially dependent on interactions with the native rhizosphere microbiota. We show that bacterial endosymbiont-harboring FOL reshapes the rhizosphere bacterial community during pathogen infection and decreased the abundance of disease{square}suppressive bacteria, including Streptomyces sp. taxa. This inhibitory effect is mediated by bacterial endosymbiont{square}stimulated production of beauvericin (an antibacterial cyclic hexadepsipeptide) by FOL. Together, our findings reveal a tripartite interaction in which a fungal pathogen leverages its bacterial endosymbiont to weaken rhizosphere microbiome{square}based disease suppression by inhibiting plant-protective bacterial taxa. This work highlights how cross{square}kingdom symbioses can modulate pathogen ecology and virulence in soil environments.

microbiology↗