Biology subjects

Yu, Z.

Publications and source records attributed to Yu, Z..

At least 109 records · Page 6Linked to original sources

bioRxiv2025

Gut Microbiota Dysbiosis and Short-chain Fatty Acid alterations in Pediatric New-Onset Type 1 diabetes with Ketoacidosis

ObjectiveDiabetic ketoacidosis (DKA) stands as the most common acute hyperglycaemic complication with a relatively high mortality rate among children diagnosed with type 1 diabetes (T1D). Previous studies have shed light on gut dysbiosis in newly-onset T1D children, but research targeting the gut microbiota and microbial metabolites in patients with DKA remains scarce. MethodsShotgun metagenomic analysis was conducted on fecal samples from 96 newly diagnosed T1D children, including 32 patients who presented with DKA upon admission. Short-chain fatty acids (SCFAs) were quantified using gas chromatography/mass spectrometry (GC/MS). Comparative analyses were performed on the gut microbiome and SCFA levels between DKA and non-DKA subgroups. ResultsGut microbiota composition differed between DKA and non-DKA groups. While alpha-diversity remained consistent, beta-diversity showed marked differences between the two groups, and the phyla Firmicutes and Bacteroidetes exhibited predominance in the DKA group. DKA was associated with increased potential pathogens and a significant depletion of SCFA-producing genera (Anaerobutyricum, Dialister, Ruminococcus, Roseburia, Dorea, and Butyricicoccus), accompanied by reduced SCFA levels. Further correlation analysis revealed strong associations between SCFAs, SCFA-producing microbiota, and disease severity indicators. Propionic acid and butyric acid mediated the association between SCFA-producing bacteria and DKA. ConclusionThis study first reported reductions in SCFAs and SCFA-producing bacteria in the fecal samples of T1D children at DKA onset. Our findings highlight SCFAs and their producers as protective biomarkers of DKA and suggest that targeting gut microbiota and SCFA metabolism may offer a potential therapeutic strategy for DKA. ImportanceWhile gut microbiota imbalances have been linked to type 1 diabetes (T1D), little is known about the role of gut microbes and their metabolites during acute diabetic ketoacidosis (DKA), an acute and life-threatening complication. This study examined the gut microbiome and levels of short-chain fatty acids (SCFAs) in fecal samples from children with new-onset T1D with and without DKA. DKA was characterized by distinct microbial signatures, including increased potential pathogens, a loss of beneficial SCFA-producing bacteria, and consistently lower SCFA levels. These changes closely correlated with indicators of disease severity, and further mediation analysis revealed the role of SCFAs in linking gut microbiota to DKA. These findings suggest that SCFAs and their producers may serve as protective markers for DKA, highlighting the potential of microbiome-based interventions to improve outcomes in children facing this critical complication.

microbiology↗

bioRxiv2025

An unrecognized and crucial role of chloroplast division in leaf variegation in Arabidopsis thaliana

Leaf variegation provides a valuable model for elucidating the molecular mechanisms that underlie chloroplast biogenesis, including both chloroplast development and division. While there has been notable advancement in comprehending the role of the Arabidopsis Yellow Variegated2 (VAR2), which encodes the AtFtsH2 subunit of the thylakoid FtsH metalloprotease complex, the exact mechanism governing leaf variegation formation remains elusive. In this study, time-course microscopy analyses of chloroplast number per cell from the first leaf pair revealed that most var2 cells lack chloroplasts. Interestingly, chloroplast biogenesis in the green sector of var2 leaves is significantly delayed but prolonged compared to the wild type (WT), leading to increased heterogeneity in chloroplast number and size among cells. These results suggest that chloroplast biogenesis in the green sector can partially compensate for the defect in the white sector. Additionally, approximately 15% of cells in var2 white sectors are devoid of plastids, and chloroplast number in guard cells is significantly lower in var2 than in WT, indicating that VAR2 mutations impair chloroplast division. Consistently, the var2 phenotype is exacerbated by knocking out key plastid division genes, including Paralog of Accumulation and Replication of Chloroplasts6 (PARC6) and Plastid Division1 (PDV1), but is fully rescued by overexpressing PDV1 or PDV2. Furthermore, VAR2 mutations inhibit etioplast development, while accelerating chloroplast biogenesis by knocking out Constitutively Photomorphogenic1 (COP1) rescues the var2 phenotype. Together, our findings reveal a novel role for VAR2 in plastid division and its essential function in chloroplast biogenesis, providing new insights into the mechanism underlying variegated leaf formation.

plant biology↗

bioRxiv2025

Learning interpretable representation for context-specific transcription regulatory networks using a foundation model

Gene expression is shaped by transcription regulatory networks (TRNs), where transcription regulators interact within regulatory elements in a context-specific manner. Despite significant efforts, understanding the intricate interactions of transcription regulators across different genomic regions and cell types remains a major challenge, largely due to data sparsity. Here, we introduce ChromBERT, a foundation model pre-trained on large-scale human ChIP-seq datasets. ChromBERT effectively captures the interaction syntax of approximately one thousand transcription regulators across diverse genomic contexts, generating interpretable representations of context-specific TRNs and their constituent regulators. Fine-tuned for various transcription regulation tasks, ChromBERT demonstrates superior performance in imputing previously unseen cistromes and modeling regulatory effects and dynamics in specific cell types. Notably, ChromBERT adapts its TRN representations for cell-type-specific tasks, providing deep insights into the roles of transcription regulators underlying observed regulatory effects or dynamics without requiring cell-type-specific genomic data for each regulator. By overcoming the limitations of sparse transcription regulator data, ChromBERT significantly enhances our ability to model, interpret, and predict transcriptional regulation across diverse biological contexts.

bioinformatics↗

bioRxiv2025

Highly heterozygous Citrus changshan-huyou Y. B. Chang originated from ancient hybridization between mandarin and pummelo and displayed distinct tissue-specific allelic imbalance

The genus Citrus is characterized by a reticulate evolutionary history with frequent hybridization, making it an intriguing subject for genome evolution investigation. Citrus changshan Y. B. Chang (Huyou) is a unique landrace first discovered in Zhejiang Province, China with premium fruit quality. The evolutionary origin of Huyou has puzzled local botanists and growers. Here, we sequenced a 120-years-old "ancestral tree" of Huyou using PacBio long read and Hi-C sequencing and assembled 2 high-quality haplotype-resolved genomes HY1 and HY2. Huyou displayed a genome heterozygosity level at 3.07%, among the highest in published citrus genomes. Using a k-mer-based tracing approach, we explicitly resolved that HY1 genome contained 87.8% mandarin, 7.3% pummelo, 0.2% citron origin, whereas HY2 had 85.0% pummelo, 2.9% mandarin, 0.3% citron, implying a hybridization event between mandarin and pummelo. Phylogeny dating showed that HY1 (2.0 Mya) and HY2 (2.18 Mya) had diverged earlier than the split of Citrus clementina and Citrus reticulata, and the split of Citrus grandis and Citrus maxima, respectively. We observed clear chromosomal recombination on chr8 and chr9 in HY1, which may have occurred after the ancestral hybridization. Further transcriptome analyses in 6 tissues revealed a strong allelic dominance of HY2 over HY1 in root tissue and moderately in stem, leaf, flower, and fruits. KEGG enrichment analyses revealed that genes related to antioxidants biosynthesis and lipid metabolisms were most significantly affected by allelic imbalance. This first report of allelic imbalance in citrus species support Huyou as an interesting model to investigate genome evolution following distant hybridization.

evolutionary biology↗

bioRxiv2025

AI-Driven Efficient De Novo design of GLP-1RAs with Extended Half-Life and Enhanced Efficacy

Peptide drug has revolutionized modern therapeutics, offering novel treatment avenues for various diseases. However, low efficacy, time consumption, and high cost hinder peptide drug design and discovery. We present an efficientin approach that integrates deep learning-based protein design with efficient functional screening, enabling the rapid design of biotechnologically important peptides with improved the stability and efficacy. We designed 10,000 de novo GLP-1 receptor agonists (GLP-1RAs), of which 60 met the stability, efficacy, and diversity criteria in the vitral functional screening. In vitro validation revealed a 62% success rate, while in vivo experiment demonstrated that two designed GLP-1RAs exhibited significantly extended half-lives, approximately three times longer than Semaglutide. In diabetic and obesity mouse models, the most competitive candidate showed superior therapeutic efficacy over Semaglutide. Our AI driven peptide design pipline integrats protein design, functional screening, and experiment validation, reducing the number of iterations required to find novel peptide candidates. The entire process, from design to screening, can be completed in a single cycle within two weeks.

bioengineering↗

bioRxiv2025

New Insights into Microbial Nitrogen Utilization in the Rumen Enabled by Genome-Resolved Multi-Omics

Optimizing nitrogen (N) utilization in ruminant production systems holds both economic and environmental significance. However, traditional paradigms of N metabolism, derived primarily from well-studied model rumen bacteria, cannot fully capture the diverse and complex N metabolic dynamics within the rumen ecosystem. To address this gap, we utilized comparative genomics and genome-resolved multi-omics analyses using a curated set of microbial genomes to investigate N assimilation and regulation in rumen microbes. We discovered that canonical mechanism of ammonia assimilation and regulation, such as the glutamine synthetase (GS)/glutamate synthase (GOGAT) pathways and its regulatory proteins, are absent in the genomes of many key and predominant rumen microbes, which likely utilize alternative pathways for ammonia assimilation. These findings challenge the applicability of E. coli-based N regulation models to rumen bacteria. We further linked polysaccharide utilization and ammonia assimilation across hundreds of rumen microbial species. Furthermore, we identified specific microbial species involved in ureolysis and denitrification, as well as phages carrying auxiliary metabolic genes that perform N assimilation. Using an animal trial involving 11 pairs of lamb twins in a crossover design, we demonstrated that dietary crude protein (CP) concentrations had minimal impact on rumen microbiome composition and expression of N assimilation genes. Instead, shifts in concentrate levels triggered alterations in N assimilation, including increased expression of amino acid biosynthesis pathways. These findings indicate a nuanced, species-specific microbial response to dietary interventions, highlighting the limitations of traditional N metabolism models applied to rumen microbes and the need for more granular studies of rumen microbial ecosystems.

microbiology↗

bioRxiv2025

Deep Learning-Based High-Resolution Time Inference for Deciphering Dynamic Gene Regulation from Fixed Embryos

Embryo development is driven by the spatiotemporal dynamics of complex gene regulatory networks. Uncovering these dynamics requires simultaneous tracking of multiple fluctuating molecular species over time, which exceeds the capabilities of traditional live-imaging approaches. Fixed-embryo imaging offers the necessary sensitivity and capacity but lacks temporal resolution. Here, we present a multi-scale ensemble deep learning approach to precisely infer absolute developmental time with 1-minute resolution from nuclear morphology in fixed Drosophila embryo images. Applying this approach to quantitative imaging of fixed wild-type embryos, we resolve the spatiotemporal regulation of the endogenous segmentation gene Kruppel (Kr) by multiple transcription factors (TFs) during early development without genetic modification. Integrating a time-resolved theoretical model of single-molecule mRNA statistics, we further uncover the unsteady-state bursty kinetics of the endogenous segmentation gene, hunchback (hb), driven by dynamic TF binding. Our method provides a versatile framework for deciphering complex gene network dynamics in genetically unmodified organisms.

biophysics↗

bioRxiv2025

How the dynamic interplay of cortico-basal ganglia-thalamic pathways shapes the time course of deliberation and commitment

Although the cortico-basal ganglia-thalamic (CBGT) network is identified as a central circuit for decision-making, the dynamic interplay of multiple control pathways within this network in shaping decision trajectories remains poorly understood. Here we develop and apply a novel computational framework -- CLAW (Circuit Logic Assessed via Walks) -- for tracing the instantaneous flow of neural activity as it progresses through CBGT networks engaged in a virtual decision-making task. Our CLAW analysis reveals that the complex dynamics of network activity is functionally dissectible into two critical phases: deliberation and commitment. These two phases are governed by distinct contributions of underlying CBGT pathways, with indirect and pallidostriatal pathways influencing deliberation, while the direct pathway drives action commitment. We translate CBGT dynamics into the evolution of decision-related policies, based on three previously identified control ensembles (responsiveness, pliancy, and choice) that encapsulate the relationship between CBGT activity and the evidence accumulation process. Our results demonstrate two contrasting strategies for decision-making. Fast decisions, with direct pathway dominance, feature an early response in both boundary height and drift rate, leading to a rapid collapse of decision boundaries and a clear directional bias. In contrast, slow decisions, driven by indirect and pallidostriatal pathway dominance, involve delayed changes in both decision policy parameters, allowing for an extended period of deliberation before commitment to an action. These analyses provide important insights into how the CBGT circuitry can be tuned to adopt various decision strategies and how the decision-making process unfolds within each regime. Author summaryWe investigate how the cortico-basal ganglia-thalamic (CBGT) network coordinates decision-making through its interconnected pathways. Using a novel Circuit Logic Assessed via Walks (CLAW) framework, we trace instantaneous neural activity through virtual CBGT networks as they engage in forced choice decisions. This analysis uncovers two key phases of a decision: deliberation, shaped by the indirect and pallidostriatal pathways, and commitment, driven by the direct pathway. We also demonstrate that CBGT activity supports two distinct decision styles: fast decisions involve an early decision boundary collapse and strong directional preference, while slow decisions feature minimal changes during an extended deliberation phase. These findings reveal the dynamic mechanisms within the CBGT network that underlie the different decision processes and how these can be tuned to adapt decision-making across varying demands and contexts.

neuroscience↗

bioRxiv2025

Nucleosome binding relinquishes the association of the BAH domain of Orc1 with Sir1

Mating-type switching in S. cerevisiae requires silencing of the homothallic mating (HM) loci through formation of position-dependent, gene-independent repressive chromatin domains, resembling heterochromatic regions in higher eukaryotes. Genetic and biochemical studies have identified cis-acting DNA elements, called silencers, and trans-acting protein factors important for the establishment and maintenance of the silent chromatin. Yet, the molecular mechanism governing the position-dependence of gene silencing is not fully understood. Here we report that the BAH domain of Orc1, which is responsible for recruiting Sir1 to the Orc1-bound silencers, ceases to bind Sir1 in the presence of nucleosome. This finding suggests a unified role of sensing the chromatin environment by Orc1s BAH domain in transcriptional silencing and specification of replication origins. We further dissected the structural determinants of the BAH domain required for binding Sir1. These results expanded the understanding of Orc1s functions in epigenetic silencing of the HM loci.

molecular biology↗

bioRxiv2025

Manually weighted taxonomy classifiers improve species-specific rumen microbiome analysis compared to unweighted or average weighted taxonomy classifiers

Previous research has demonstrated that applying taxonomic weights to shotgun metagenomic data can improve species identification in 16S rRNA gene-based microbiome analysis. However, such an approach does not allow for accurate analysis of samples collected from less studied habitats, such as rumen. In the present study, we developed a method to incorporate taxonomic weights based on relative abundance of species identified from shotgun sequencing and amplicon sequencing data derived from rumen. Using this weighting method, we evaluated latest versions of five prominent databases--SILVA, Greengenes2, RDP, NCBI RefSeq, and GTDB--against the BLAST 16S rRNA database, assessing classification counts, fully classified ratios (proportion of ASVs classified to a known genus and species), and error rates. Our results revealed that the use of the weighting method significantly improved both classification counts and fully classified ratios, along with a substantial (P < 0.05) reduction in error rates compared to unweighted taxonomy classifier. While GG2 and SILVA struggled with accurate classification at the species level owing to their inherent database characteristics, GTDB consistently improved all metrics using the manually weighted taxonomy classifier, achieving up to an 8% error rate reduction at the species level. NCBI RefSeq and RDP also exhibited remarkable improvement in the classification counts and fully classified ratios, along with substantial error rate reductions by up to 47% at the species level. These findings demonstrate that amplicon sequencing datasets can enhance rumen microbiome analyses through effective weighting methods. While SILVA is commonly used in metataxonomic analyses of the rumen microbiome, we recommend NCBI RefSeq for species-level classification due to its superior accuracy and minimal ambiguous classification (e.g., "uncultured" or "sp.") in future metataxonomic studies.

bioinformatics↗

bioRxiv2025

ProDualNet: Dual-Target Protein Sequence Design Method Based on Protein Language Model and Structure Model

Proteins typically interact with multiple partners to regulate biological processes, and peptide drugs targeting multiple receptors have shown strong therapeutic potential, emphasizing the need for multi-target strategies in protein design. However, most current protein sequence design methods focus on interactions with a single receptor, often neglecting the complexity of designing proteins that can bind to two distinct receptors. We introduced ProDualNet, a novel approach for designing dual-target protein sequences by integrating sequence-structure information from two distinct receptors. ProDualNet used a heterogeneous graph network for pretraining and combines noise-augmented single-target data with real dual-target data for fine-tuning. This approach addressed the challenge of limited dual-target protein experimental structures. The efficacy of ProDualNet has been validated across multiple test sets, demonstrating better recovery and success rates compared to other multi-state design methods. In silico evaluation of cases like dual-target allosteric binding and non-overlapping interface binding highlights its potential for designing dual-target binding proteins. Furthermore, we validated ProDualNets ability to model the relationships between sequences, structures, and functions by zero-shot prediction tasks, including dual-target protein functional effects and mutant functional effects.

bioinformatics↗

bioRxiv2025

Phosphorylation State Dictates Bacterial Stressosome Assembly and Function

Bacterial pathogens rely on their ability to sense and respond to environmental stressors to survive and maintain virulence. The stressosome, a 1.8-megadalton nanomachine, serves as a critical sensor and regulator of the general stress response. It is composed of multiple copies of three proteins RsbR, RsbS, and the kinase RsbT which together orchestrate activation of downstream stress adaptation pathways. Using cryo-electron microscopy, we solved the atomic structure of five Listeria monocytogenes stressosomes, capturing structural mimics of the transition between inactive and activated states using phosphomimetic and phosphodeficient mutants. Our findings reveal that phosphorylation at specific residues T175 and T209 on RsbR, and S56 on RsbS dictates stressosome assembly, stoichiometry, and activation. Specifically, phosphorylation at T175 primes the stressosome for activation, while S56 phosphorylation destabilizes the core, triggering the release of RsbT to propagate the stress response. In contrast, phosphorylation at T209 modulates stressosome composition and appears to fine-tune the intensity of the stress response. Functional analyses reveal that phosphomimetic mutants (T209E, S56D) resist oxidative stress but lose virulence in host cell model, while phosphodeficient mutants (T175A, S56A) are stress-sensitive but retain virulence. These findings establish phosphorylation as a central regulatory switch linking structural dynamics to bacterial adaptation and pathogenesis, highlighting potential targets for antimicrobial intervention.

microbiology↗

bioRxiv2025

Precise metabolic dependencies of cancer through deep learning and validations

Cancer cells exhibit metabolic reprogramming to sustain proliferation, creating metabolic vulnerabilities absent in normal cells. While prior studies identified specific metabolic dependencies, systematic insights remain limited. Here we build a graph deep learning based metabolic vulnerability prediction model "DeepMeta", which can accurately predict the dependent metabolic genes for cancer samples based on transcriptome and metabolic network information. The performance of DeepMeta has been extensively validated with independent datasets. The metabolic vulnerability of "undruggable" cancer driving alterations have been systematically explored using the cancer genome atlas (TCGA) dataset. Notably, CTNNB1 T41A activating mutations showed experimentally confirmed vulnerability to purine/pyrimidine metabolism inhibition. TCGA patients with the predicted pyrimidine metabolism dependency show a dramatically improved clinical response to chemotherapeutic drugs that block this pyrimidine metabolism pathway. This study systematically uncovers the metabolic dependency of cancer cells, and provides metabolic targets for cancers driven by genetic alterations that are originally undruggable on their own.

cancer biology↗

bioRxiv2025

Endurance Exercise Ameliorates Aging-Related Bradyarrhythmia in Drosophila Resulting from miR-283 Knockdown in LNvs

MicroRNAs (miRNAs) are crucial in regulating cardiac aging and related diseases, yet few functional miRNAs have been identified. Prior studies showed miR-216a upregulation in heart failure patients, but its impact on aging hearts is unknown. Our study revealed systemic miR-283 overexpression or knockdown caused age-related bradycardia, mimicking human bradyarrhythmia. Importantly, we found that knockdown of miR-283 in ventral-lateral neurons (LNvs), rather than in the heart, led to the occurrence of bradyarrhythmia, which was mainly caused by the upregulation of miR-283 expression in the whole brain and heart. The gene of clockwork orange (cwo) may mediate miR-283s effect on heart rhythm. Additionally, to investigate the miRNA regulatory mechanism underlying exercise-induced delay in cardiac aging, we conducted a three-week endurance exercise program on miR-283 knockdown flies in LNvs. We found that exercise significantly downregulated the accumulation of miR-283 in the brain and myocardium caused by aging or miR-283 knockdown in LNvs, improved the structure of myocardial fibers, and effectively reduced bradyarrhythmia. Our findings provides a new perspective on distal neuromodulation and intervention in cardiac aging.

physiology↗

bioRxiv2025

A pan-gene catalogue of Asian cultivated rice

The rice genome underpins fundamental research and breeding, but the Nipponbare (japonica) reference does not fully encompass the genetic diversity of Asian rice. To address this gap, the Rice Population Reference Panel (RPRP) was developed, comprising high-quality assemblies of 16 rice cultivars to represent japonica, indica, aus, and aromatic varietal groups. The RPRP has been consistently annotated, supported by extensive experimental data and here we report the computational assignment, characterization and dissemination of stably identified pan-genes. We identified 25,178 core pan-genes shared across all cultivars, alongside cultivar-specific and family-enriched genes. Core genes exhibit higher gene expression and proteomic evidence, higher confidence protein domains and AlphaFold structures, while cultivar-specific genes were enriched for domains under selective breeding pressure, such as for disease resistance. This resource, integrated into public databases, enables researchers to explore genetic and functional diversity via a population-aware "reference guide" across rice genomes, advancing both basic and applied research.

genomics↗

bioRxiv2025

Repurposing AZD-5991 for inhibiting growth and biofilm formation of Staphylococcus aureus by disrupting the cell membrane and targeting FabI.

Staphylococcus aureus infections have already become a major threat to public health worldwide. Drug resistance and biofilm formation are two critical factors that significantly affect the efficacy of the antimicrobial treatment of S. aureus infections using conventional antibiotics. Therefore, the discovery of novel antimicrobial agents with potent antibacterial and antibiofilm activity has become a hotspot in recent years. Here, we first demonstrated the remarkable inhibitory activity of AZD-5991, a selective Mcl-1 inhibitor, against S. aureus. The MIC50 and MIC90 of AZD-5991 against S. aureus were 12.5 M, and AZD-5991 significantly inhibited the growth of S. aureus at a subinhibitory concentration of 1/2 x MIC. Besides, AZD-5991 displayed bactericidal activity and a robust capacity for inhibiting biofilm formation against S. aureus with very low drug toxicity against host cell lines. Our data demonstrated the disruption of S.aureus cell integrity by AZD-5991 through membrane permeability assays and the ingredients of the bacterial phospholipids could neutralize the antibacterial activity of AZD-5991. Moreover, whole-genome sequencing and proteomic analysis were also applied to gain insights into the possible impact of AZD-5991 on the fatty metabolism of S. aureus. Furthermore, the antibacterial activity of AZD-5991 was significantly elevated by exogenous fatty acids linoleic acid (C18:2 {Delta} 9,12) and arachidonic acid (C20:4 {Delta} 5,8,11,14). Lastly, the biolayer interferometry assay supported the direct interaction of AZD-5991 with FabI, a necessary protein related closely to bacterial growth and fatty acid metabolism. Conclusively, this study suggests that AZD-5991 could inhibit the planktonic growth and biofilm formation of S. aureus by disrupting the cell membrane and targeting FabI. AZD-5991 might be a promising new antibiotic candidate for the antimicrobial treatment of S. aureus infections resistant to traditional clinical drugs. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=84 SRC="FIGDIR/small/638150v1_ufig1.gif" ALT="Figure 1"> View larger version (18K): org.highwire.dtl.DTLVardef@658444org.highwire.dtl.DTLVardef@cea949org.highwire.dtl.DTLVardef@1cc003dorg.highwire.dtl.DTLVardef@23de22_HPS_FORMAT_FIGEXP M_FIG C_FIG

microbiology↗

bioRxiv2025

Impact of cross-validation designs on cattle behavior prediction using machine learning and deep learning models with tri-axial accelerometer data

1Assessing cattle behaviors provides insights into animal health, welfare, and productivity to support on-farm management decisions. Wearable accelerometers offer an alternative approach to traditional human evaluation, providing a more objective and efficient method for predicting cattle behavior. Random cross-validation (CV) is commonly used to evaluate behavior prediction by splitting data into training and testing sets, but it can yield inflated results when records from the same animal are included in both sets. Block CV splits data by block effects, offering a more realistic evaluation but remains underexplored for predicting multi-class imbalanced cattle behavior. Additionally, deep learning (DL) models have not been fully explored for behavior prediction compared to machine learning (ML) models. The objectives of this study were to examine the impact of CV designs on multi-class imbalanced cattle behavior prediction and to compare the performance of ML and DL models. Three ML and two DL models were used to predict the four behaviors of six beef cows from a public tri-axial accelerometer dataset, with model performance evaluated using both hold-out and leave-cow-out CV designs representing random and block CV, respectively. In hold-out CV, ML models achieved accuracies of 0.94 to 0.96 and F1 scores of 0.93 to 0.95, while DL models achieved accuracies of 0.9 to 0.92 and F1 scores of 0.89 to 0.91. In the leave-cow-out CV, ML models obtained accuracies of 0.72 to 0.82 and F1 scores of 0.64 to 0.82, whereas DL models obtained accuracies of 0.76 to 0.82 and F1 scores of 0.64 to 0.76. Generally, ML models outperformed DL models in the hold-out CV, but the multi-layer perceptron DL model demonstrated comparable or superior performance in the leave-cow-out CV. All models performed better with hold-out CV than leave-cow-out CV. Our results suggest that CV designs can affect behavior prediction performance. While a random CV produces seemingly good predictions, these results can be artificially inflated by the data partition. A block CV that strategically partitions data could be a more appropriate design.

animal behavior and cognition↗

bioRxiv2025

Multi-scale structure of chromatin condensates rationalizes phase separation and material properties

Biomolecular condensates, compartments that concentrate molecules without surrounding membranes, are integral to numerous cellular processes. The structure and interaction networks of molecules within condensates remain poorly understood. Using cryo-electron tomography and molecular dynamics simulations we elucidated the structure of phase separated chromatin condensates across scales, from individual amino acids to network architecture. We found that internucleosomal DNA linker length controls nucleosome arrangement and histone tail interactions, shaping the structure of individual chromatin molecules both within and outside condensates. This structural modulation determines the balance between intra- and intermolecular interactions, which in turn governs the molecular network, thermodynamic stability, and material properties of chromatin condensates. Mammalian nuclei contain dense clusters of nucleosomes whose non-random organization is mirrored by the reconstituted condensates. Our work explains how the structure of individual chromatin molecules ultimately determines physical properties of chromatin condensates, with implications for cellular chromatin organization.

biophysics↗