Biology subjects

Lee, B.

Publications and source records attributed to Lee, B..

At least 19 recordsLinked to original sources

bioRxiv2026

Targeting skin ageing at the source: a novel cellular reprogramming approach using natural extract combinations

Cellular senescence is a driver of skin ageing that leads to structural deterioration and aesthetic changes. This study evaluated a novel senomorphic strategy, using synergistic combinations of natural small molecules to target RNA splicing regulation via upstream ERK/AKT signaling pathways. Nature-identical synthetic molecules, and corresponding extracts were assessed for cytotoxicity, impact on senescent cell load, proliferation, DNA damage load and SASP markers in naturally aged human primary skin cells. A topical, simple formulation was evaluated for impact on senescence and skin structure using 3D human ex vivo full-thickness skin explants exposed to 1.5 J/cm2 UVC irradiation. Combination of H. lupulus and G. biloba extracts produced a dose-dependent senomorphic impact of up to 45% in aged dermal fibroblasts and 10% in aged human primary keratinocytes (p [≤] 0.0001 & 0.001), without inducing cell proliferation. A 50% decrease in DNA damage load was also observed (p [≤] 0.0001). Qualitatively, the simple formulation mitigated UV-induced epidermal hyperplasia, restored collagen fiber organization, and reduced markers of senescence and inflammation. Targeting cellular senescence is an emerging traction point to address the root biological causes of skin ageing. Targeting the splicing dysregulation/senescence axis using combinations of natural plant extracts may provide a novel mode of action to reprogram senescent cells for repair and protection of skin structure and function.

cell biology↗

bioRxiv2026

Limits of marker information transfer]{Limits of marker information transfer in cross-panel cytometry imputation

Marker imputation can bridge cytometry datasets acquired with different panels and platforms. Its validity depends on whether common markers retain information about unmeasured target markers and whether those relationships transfer across assay contexts. We evaluated these limits in two paired flow cytometry-CyTOF cohorts comprising 134 patients and 257 specimens. A patient-held-out framework distinguished within-platform same-cell reconstruction from cross-platform distribution translation without cell correspondence. We compared conventional machine-learning and deep generative estimators with median imputation and used permutation controls to estimate the contributions of patient matching and source-exclusive markers. Evaluation covered marker values, specimen distributions, phenotype utility, and directional predictive relationships among markers. Within platforms, exact k-nearest neighbors reduced mean absolute error by 12.8-35.0% relative to median imputation. However, the median-based reconstructions were contracted, with a median ratio of predicted to observed interquartile range below 1 in every context and evaluable population. Patient matching improved cross-platform distribution recovery, whereas adding source-exclusive markers or higher-capacity estimators did not consistently produce net gains. In the independently annotated acute myeloid leukemia (AML) cohort, reconstructed markers largely preserved broad-lineage classification but did not recover fine-phenotype information available through direct measurement. In AML-CyTOF, the measured full panel outperformed the reconstructed representation by 10.25 percentage points for fine phenotypes. Predictive relationships among markers were context-specific. Most targets were better explained by several markers than by one. Marker selection based on these relationships showed a trade-off between panel size and pairwise coverage. Paired samples and imputation can therefore connect selected marker distributions across panels but should not be treated as substitutes for direct measurement. Cross-panel cytometry should be evaluated with patient-held-out tests matched to the resolution of available ground truth and the intended downstream use.

bioinformatics↗

bioRxiv2026

Human single-neuron correlates of somatosensation

Understanding the neural correlates of consciousness poses a central challenge in neuroscience, with longstanding debate over which cortical areas underlie subjective sensory perception. In the tactile domain, posterior parietal cortex (PPC) and primary somatosensory cortex (S1) are candidate cortical substrates for neural correlates of consciousness due to their crucial roles in somatosensory processing. Yet whether and how these regions contribute to conscious tactile perception remains unclear. Here, to address this issue, we leveraged chronic intracortical recordings to simultaneously monitor large ensembles of single neurons in grasp related cortical areas from three participants with tetraplegia while natural vibrotactile stimuli were delivered to a sensate region of the upper limb. We demonstrate that an orchestration of neural activity emerged across the grasp-related parietal-frontal network at near-threshold perception. Within this distributed network, we then focused on PPC and S1. While S1 responses closely tracked the amplitude of the physical stimulus, PPC activity covaried more strongly with awareness and confidence. This single-neuron evidence reveals PPC and S1 have complementary neural codes with PPC transforming physical stimulation into conscious perception whereas S1 is more sensitive to stimulus amplitude.

neuroscience↗

bioRxiv2026

Multi-omic profiling reveals impaired luteal resolution and CHK2-regulated inflammation in ovarian aging

The ovary ages earlier than most other organs, but the molecular mechanisms that drive this decline remain unclear. To gain more insight into this process, we profiled ovaries from young and reproductively aged mice after synchronized ovulation and found that aging was marked by the accumulation of luteal cells with increased steroidogenic, antigen presentation and inflammatory gene expression. DNA methylation was globally preserved in aged ovaries but showed focal age-associated changes at gene regulatory regions. Protein profiling identified an increase in checkpoint kinase 2 (CHK2), a DNA damage response protein encoded by the CHEK2 gene associated with the timing of natural menopause in humans. CHK2 and DNA damage markers were increased in aged mice and human ovarian somatic cells. Transient CHK2 inhibition in aged mice increased the proportion of mature oocytes and blastocyst formation and reversed inflammatory signatures in granulosa cells. These findings identify defective luteal resolution and inflammation as features of ovarian aging and suggest that CHK2 induces a reversible age-associated inflammatory signature. Our results show that reproductive decline may remain pharmacologically modifiable during ovarian aging via regulation of the inflammatory response.

molecular biology↗

bioRxiv2026

Neural encoding of grasp and object properties in the posterior parietal and motor cortices of the cortical grasping network in tetraplegic humans

The cortical grasping network (CGN) is responsible for the ability to physically interact with the world around us using our hands. For patients with motor impairments due to neurogenerative disease or traumatic injury, brain-machine interfaces (BMIs) offer a potential pathway towards restoration of dexterous hand control via recording neural activity throughout the CGN. However, grasping objects with robotic BMI devices has proved challenging, when utilizing neural signals from only motor cortex (MC, a subregion of the CGN). It is currently unclear how the presence of an object might compromise BMI performance; thus this work explores the interactive neural representation of grasp and objects throughout the CGN. Three tetraplegic human participants performed grasp motor imagery during imagined object manipulation while we recorded neural activity from the supramarginal gyrus (SMG), anterior intraparietal cortex (AIP), motor cortex (MC), and primary somatosensory cortex (S1). All regions within the CGN represented whole hand configuration of imagined grasps during motor planning and imaged execution. Additionally, grasp-related neural activity in each region was modulated by context (motor planning vs. imagined execution and object present vs. not present). SMG and AIP represented object shape during motor planning. PPC encoded both grasp and object properties simultaneously from mostly unique subpopulations of neurons. This separability in higher cortical regions could be a critical for stable BMI grasp performance.

neuroscience↗

bioRxiv2026

Hyperactivation of the AXL-ICD/SIRT2 axis by Amyloid-β impairs astrocytic autophagic flux and exacerbates neuroinflammation

Autophagy dysfunction and neuroinflammation are central to Alzheimer's disease (AD), yet how extracellular amyloid-{beta} (A{beta}) couples to impaired autophagic flux and heightened neuroinflammation remains unknown. Here, we identify the TAM receptor AXL as a molecular transducer that couples A{beta} sensing to the regulation of autophagy and neuroinflammation in astrocytes. A{beta} induces {gamma}-secretase-dependent cleavage of AXL, generating a nuclear intracellular domain (AXL-ICD) that forms phase-separated condensates and activates autophagy gene transcription through SIRT2-mediated recruitment of the RUVBL1/2-INO80 chromatin-remodeling complex. This axis is activated in astrocytes of postmortem AD brains. Concurrently, AXL-ICD binds to the SIRT2 catalytic domain and suppresses its deacetylase activity, increasing -tubulin acetylation and altering microtubule dynamics. While moderate AXL-ICD levels promote autophagic flux, excessive elevation paradoxically triggers microtubule hyperstabilization, thereby impairing autophagosome-lysosome fusion and causing pathological accumulation of autophagosomes and H2O2. The inhibitory peptide AxSBiP disrupts the AXL-ICD/SIRT2 interaction, restores autophagic flux, reduces plaque burden, and normalizes A{beta}-induced H2O2 production and astrogliosis in APP/PS1 mice. We propose the AXL-ICD/SIRT2 axis as an effective therapeutic target to reduce A{beta} burden and neuroinflammation in AD

neuroscience↗

bioRxiv2026

CytoGate-Bench: an LLM benchmark for cross-panel cell gating in cytometry

In cytometry, the workhorse single-cell technology of clinical immunology, every study defines its own antibody panel and cell-type vocabulary, so a classifier trained on one cannot annotate the next. Immunologists instead annotate by manual gating, splitting one parent population at a time on a two-marker plot, down an expert-defined hierarchy. We introduce CytoGate-Bench, a benchmark that reformulates this per-step procedure as a zero-shot, panel-agnostic task for large language models. It comprises 23,646 expert-annotated instances re-curated from 11 public flow- and mass-cytometry cohorts spanning eight marker panels. Across six open- and closed-weight backbones, the strongest formulation draws one rectangular gate per candidate and falls within the range of trained, panel-specialized baselines. It degrades less under distribution shift. Walking the hierarchy stepwise outperforms predicting every cell type at once. Ablations trace the signal to the data distribution shape and curated marker priors. However, adding vision or a self-verification loop systematically tightens gates.

bioinformatics↗

bioRxiv2026

Programmable De Novo Design of Mesoporous Protein Crystal Frameworks

Three-dimensional protein crystals are ordered, porous macroscopic materials with potential applications in catalysis, biosensing, and biomedicine. However, most protein crystals are obtained by empirical screening, providing limited control over the lattice architecture, pore geometry or component composition that determine material function. Here, we present a modular strategy for the programmable design of highly porous, framework-like protein crystals using predefined protein-protein interactions. This strategy yielded over 30 distinct protein crystals, including single-component and multicomponent P213 and I213 lattices that grow to over 100 micrometers in size. Small-angle X-ray scattering and electron microscopy showed close agreement between experimental lattices and computational models. RFdiffusion-guided design generated isomorphous variants with matched lattice parameters, enabling coherent protein crystal alloys, epitaxial core-shell growth and reversible shell assembly. The designed crystals exhibit tunable mesoporous architectures, with limiting apertures of 2-18 nm, and support genetically encoded incorporation of fluorescent protein guests. These results establish a general route to programmable lattice engineering of protein crystals and position them as genetically encoded, compositionally tunable mesoporous materials.

synthetic biology↗

bioRxiv2026

scDIVA: semi-supervised integration and fine-grained annotation of tumor-immune single-cell atlases

Single-cell atlases of the tumor-immune microenvironment have defined numerous fine-grained immune cell states, but each study uses its own nomenclature and procedure for annotating cell types. Transferring annotations from a reference atlas to a query dataset is complicated by both batch effects and by the presence of query populations that the reference does not contain. Here we present scDIVA, a semi-supervised deep generative model that adapts the Domain Invariant Variational Autoencoder to scRNA-seq for fine-grained tumor-immune label transfer. Three encoders disentangle each cells expression profile into separate latent subspaces for cell type, batch, and residual variation; a single decoder reconstructs the cells expression profile from all three latent embeddings, and auxiliary classifiers on the cell type and batch embeddings encourage each encoder to capture the respective source of variation; scDIVAs cell type embeddings are batch-invariant by construction rather than through explicit or adversarial correction. Benchmarked against four established reference-mapping approaches--Harmony/Symphony, scANVI with scArches, scPoli with scArches, and Seurat with label transfer--across six tumor-immune atlases spanning five cancer types, scDIVA achieved the highest mean macro-F1 in five of six atlases and the highest biological conservation scores. To detect query-enriched populations, we adapted the Milo differential abundance (DA) framework, added a directional test, corrected spatialFDR weighting, and parallelized neighborhood distances for atlas-scale data, and applied it to scDIVAs cell type embeddings with reference- versus-query membership as the condition. This design correctly identified cell types held out from the reference as OOR and flagged exhausted CD8 T cells from a tumor-immune atlas as OOR relative to a healthy pan-tissue immune reference; conversely, this procedure confirmed a conserved immune landscape between two independent colorectal cancer cohorts. scDIVA thus couples fine-grained annotation and integration with an FDR-controlled test for states the reference lacks, solving both tasks required for accurate label transfer in tumor-immune scRNA-seq atlases.

bioinformatics↗

bioRxiv2026

Whole-brain modeling of dynamic causal circuits in human cognition using amortized variational inference

Understanding dynamic mechanisms underlying cognition remains a major challenge in human neuroscience. Here, we develop, validate, and apply Multivariate Dynamical Systems Identification with Amortized Variational Inference (MDSI-AVI), a novel computational framework designed to address critical challenges in capturing asymmetric, context-dependent, whole-brain directed interactions while accounting for regional hemodynamic response variability in fMRI data. MDSI-AVI leverages simulation-based inference through forward and reverse variational inference to address the limitations of conventional variational methods in high-dimensional settings. By averaging over uncertainty in hemodynamic response parameters using forward simulation, MDSI-AVI provides well-calibrated posteriors of directed connectivity that scale efficiently to networks with hundreds of nodes. Applied to Human Connectome Project data (N=728), MDSI-AVI reveals new insights into working memory mechanisms, identifying the dorsal anterior insula as a critical hub influencing activity at the whole-brain level. We demonstrate task-dependent modulation of causal influences, where the salience network drives frontoparietal network activity, which differentially influences the default mode and sensorimotor networks depending on working memory load. These whole-brain causal interactions distinguish task conditions with high accuracy and predict working memory performance. Our framework demonstrates reproducible results across whole-brain parcellations, establishing MDSI-AVI as a robust tool for advancing our understanding of circuit dynamics in cognition and disease.

neuroscience↗

bioRxiv2026

Boltz-Perturb: Improving Diversity and Accuracy in Protein-Ligand Co-Folding through Training-Free Conditioning Perturbation

Protein-ligand co-folding models hold promise in structure-based drug discovery and small molecule interaction prediction, but often fail in predicting correct small molecule binding poses. We present Boltz-Perturb, a framework for addressing this through perturbing model conditioning signals during model inference, and show that such perturbations improve correct ligand binding mode predictions. We first show with true-coordinate injection experiments that the models learned energy landscape contains correct binding-mode basins, allowing us to reframe the problem as one of sampling deficiency. We then introduce two inference-time perturbation strategies, Token Bias Perturbation (TBP) and Token Conditioning Perturbation (TCP), which increase exploration of alternative binding poses. Across diverse protein-ligand systems, TCP improves top-20 oracle success rates by 2.6 to 7.8 fold. Boltz-Perturb attains higher oracle success rates compared to the Boltz-2 high diffusion temperature variant while requiring over 75% less compute. To our knowledge, this is the first systematic perturbation analysis of a co-folding architecture for small-molecule binding mode diversity. We demonstrate that inference-time perturbations can unlock latent structural diversity in generative co-folding models and improve protein-ligand predictions without costly retraining.

biophysics↗

bioRxiv2026

Sequential intra-articular HCAd-NFκB-IL-1Ra delivery improves therapeutic efficacy in post-traumatic osteoarthritis

Osteoarthritis is the most common joint disease for which disease-modifying therapies remain unavailable. Intra-articular gene delivery of interleukin-1 receptor antagonist (IL-1Ra) using high-capacity adenovirus (HCAd) has shown therapeutic promise; however, the duration of therapeutic benefit and the feasibility of repeat dosing under anti-adenoviral immunity remain unresolved. Using the murine anterior cruciate ligament transection model of osteoarthritis, we show that a single intraarticular injection of HCAd5-NF{kappa}B-IL-1Ra provides structural preservation and functional improvement in early-stage osteoarthritis but fails to sustain cartilage protection as disease progresses. However, HCAd5 transduction following repeated treatment is limited due to pre-existing immunity against this serotype. Notably, exchanging serotypes for repeated treatments effectively restores vector transduction and transgene expression in both healthy and osteoarthritic joints. Leveraging this strategy, we demonstrate that sequential intra-articular HCAd-NF{kappa}B-IL-1Ra administration does not further improve pain or motor function compared to the initial treatment, but preserves cartilage as assessed by histopathology and phase-contrast CT, irrespective of serotype. These findings establish HCAd serotype switching as a feasible approach to overcome immune barriers to repeat intra-articular gene therapy. Importantly, sequential HCAd-NF{kappa}B-IL-1Ra administration enhances therapeutic durability in post-traumatic osteoarthritis, providing a translational framework for repeat intra-articular gene delivery strategies aimed at long-term disease modification in osteoarthritis.

molecular biology↗

bioRxiv2026

Cross-Species Efficacy of Combinatorial Gene Therapy for Osteoarthritis and Correction of Neuro-Inflammatory Pain Mechanisms

Osteoarthritis is a leading cause of chronic pain and disability, which lacks disease-modifying treatment. Given the complex multi-tissue and multifactorial drivers behind disease progression, effective treatments will require simultaneously targeting several mechanisms underlying joint degeneration and pain. Here, we developed and evaluated a combinatorial gene therapy, consisting of a high-capacity adenoviral vector carrying two therapeutic genes to target distinct pathological mechanisms: inflammation (IL-1Ra) and chondrocyte health (PRG4). Intra-articular delivery of this treatment improved functional, structural, and pain outcomes in murine and equine osteoarthritis models. In addition, treatment normalized inflammatory environments in joint tissues, as well as in the dorsal root ganglia (DRG) known to harbor joint-innervating sensory neurons. Moreover, gene therapy reversed OA-induced molecular signatures of neural hyperexcitability, suggesting amelioration of peripheral sensitization. Collectively, these findings support combinatorial gene therapy as a promising treatment for osteoarthritis, while identifying neuroinflammatory signatures for correction of disease progression and pain. One Sentence SummaryA single intra-articular injection of a combinatorial gene therapy slows OA progression and reduces pain in small and large animal models.

systems biology↗

bioRxiv2026

Coexisting but dissociable place and spatial view codes in the primate hippocampus

In rodents, hippocampal activity is classically described in terms of position-anchored place cells, whereas nonhuman primate work has emphasized spatial view cells, in which activity depends on the region of space being viewed. This contrast has fueled a longstanding debate over whether primate and rodent spatial codes are biologically distinct. We recorded single units in the hippocampus of macaques as they navigated linear tracks in two visually distinct contexts, intermittently making context-dependent object choices. Using a generalized linear model to dissociate position coding from spatial view coding, we identified three coexisting populations in the macaque hippocampus: Position cells, View cells, and Conjunctive cells encoding both variables. Position cells were prevalent and showed rodent-like properties--localized firing, directional tuning, and contextual remapping--even after accounting for visual sampling, whereas View cells stably encoded allocentric visual space. Population decoding confirmed this dissociation at the ensemble level. Robust position codes thus coexist with, yet remain dissociable from, view codes in the primate hippocampus, suggesting that the apparent scarcity of place-like responses in earlier primate work reflects methodological and task-related differences rather than a fundamental species-level divergence.

neuroscience↗

bioRxiv2026

StructureSAFE: A structure-aware chemical language model for unified hit identification and lead optimization

Structure-based generative models (SBGMs) hold great promises for accelerating drug discovery by enabling target-aware molecular design. However, existing approaches face fundamental challenges: three-dimensional graph-based models can explicitly incorporate protein structural information but often generate chemically implausible molecules due to limited training data, while chemical language models (CLMs) produce chemically plausible molecules but struggle to effectively leverage three-dimensional structural information for structure-conditioned generation and hard to incorporate lead optimization functionality due to the nature of SMILES string. Here, we present StructureSAFE, a structure-aware chemical language model that resolves this trade-off by integrating protein structural and evolutionary encoders with the SAFE molecular representation via pretraining and finetuning training scheme, enabling both de novo hit identification and a comprehensive suite of lead optimization subtasks within a unified framework. Comprehensive benchmarking on the MolGenBench dataset demonstrates that StructureSAFE achieves state-of-the-art (SOTA) performance across multiple metrics, with particularly pronounced improvements in chemical plausibility relative to graph-based models lacking pretraining. Evaluation on a rigorously constructed held-out test set further confirms its ability to generate drug-like, synthetically accessible molecules with competitive predicted binding affinities for previously unseen targets on both hit identification and lead optimization setting. In silico case studies across four therapeutically relevant targets validate its capacity to generate chemically plausible molecules that recapitulate key binding interactions of known high-affinity ligands while proposing novel interactions for potential better affinity and exploring previously unknown regions of chemical space. Taking together, StructureSAFE represents a versatile and practical tool to provide high-quality candidate molecules for augmenting medicinal chemistry workflows in both hit identification and lead optimization campaigns.

bioinformatics↗

bioRxiv2026

Toward understanding drivers of specialized metabolism in Actinomycetota: insights from 1432 transcriptomics datasets for 132 strains

Actinomycetota are major sources of specialized metabolites with applications in drug discovery and agriculture, yet much of their biosynthetic potential remains silent under standard laboratory conditions. Limited understanding of the mechanisms controlling biosynthetic gene cluster (BGC) activation constrains metabolite discovery and production. Here, we generated 1432 RNA-seq datasets from 132 Actinomycetota strains grown in eight media to identify patterns associated with BGC expression. On average, strains expressed 44% of their encoded BGCs across the tested conditions, with more expression observed among known BGCs (61%) compared to uncharacterized BGCs (35%). Co-expression analyses revealed frequent associations between BGCs and transporters, transcriptional regulators, and proteins containing DUF397 and DUF742 domains. Targeted overexpression of candidate genes selected from BGC-associated co-expression modules increased metabolite production, with DUF397- and DUF742-containing operons showing the broadest effects by boosting the levels of several different specialized metabolites. Other genes boosted levels in a metabolite-specific manner. Together, our results support a multilayered model of BGC regulation in Actinomycetota in which BGC expression is shaped by medium composition, BGC-specific regulators, and integration of BGCs into broader transcriptional network modules. By connecting BGC expression to specific media and co-expressed genes, this study also provides a resource for selecting growth conditions and engineering specialized metabolism.

microbiology↗

bioRxiv2026

Synergistic anti-tumor activity of EGFR inhibition and C/EBPβ antagonism in GBM.

Glioblastoma (GBM) is the most prevalent primary brain cancer, with poor prognosis and limited therapeutic options available. The genetic and cellular heterogeneity characteristic of GBM contributes to poor response rates. Activating mutations of the epidermal growth factor receptor (EGFR) gene are among the most frequent alterations in GBM, occurring in roughly half of cases. Despite the prevalence of EGFR mutations, EGFR inhibition has shown limited success in GBM. The transcription factor C/EBP{beta} is a master regulator of the mesenchymal transformation in GBM, an aggressive state characterized by increased invasiveness and resistance to chemotherapy. Lucicebtide is a C/EBP{beta} antagonist peptide with demonstrated single agent activity in patients with recurrent GBM that is currently being evaluated in a clinical trial in combination with radiation and temozolomide in patients with newly-diagnosed GBM (NCT04478279), with emerging data supporting clinical activity in that setting. Here we show that in the TCGA-GBM dataset, patients with EGFR mutations display significant enrichment of a high C/EBP{beta} activity signature. Functionally, genetic inactivation of EGFR by CRISPR results in synthetic lethality in the presence of lucicebtide in GBM cell lines, and synergistic in vitro cytotoxicity and suppression of C/EBP{beta} target gene expression was observed in combination experiments with lucicebtide and EGFR inhibitors. Finally, enhanced anti-tumor activity was demonstrated in vivo in the combination setting, as combined subpharmacologic dose levels of lucicebtide and the EGFR inhibitor osimertinib potently suppressed GBM xenograft growth. These data identify EGFR and C/EBP{beta} dependencies in GBM and support lucicebtide combination with EGFR inhibitors as a potential therapeutic option for a sizable fraction of GBM patients.

cancer biology↗

bioRxiv2026

Deep learning based behavioral analysis in a neonatal rat model of hypoxic ischemic brain injury

Hypoxic-ischemic (HI) brain injury in neonates is one of the leading causes of lifelong neurological disability. Behavioral tests in preclinical rodent models are widely used to assess motor and cognitive outcomes after HI injury; however, these assays usually depend on subjective and labor-intensive manual scoring. Recent advances in markerless pose estimation offer new opportunities for automated and reproducible behavioral quantification in animal and infant recordings, but their use in neonatal HI preclinical studies remains limited. Wistar rat pups underwent HI injury using the Rice-Vannucci model at postnatal day 7 (P7). Three developmental behavioral tests included righting reflex (P8), negative geotaxis (P14), and wire hang (P16), were recorded and analyzed by both a human rater and an automated pipeline using DeepLabCut (DLC), an open source markerless pose estimation framework. Automated measurements were compared with manual scores using Intraclass Correlation Coefficients (ICC), Bland-Altman analysis, and Pearson correlation. DLC-derived measurements demonstrated strong agreement with manual scoring across all assays. ICC values were 0.929 (95% CI 0.648-0.971) for righting reflex, 0.965 (0.888-0.989) for negative geotaxis, and 0.958 (0.876-0.985) for wire hang. An automated behavioral analysis framework integrating DLC-based pose estimation with rule based quantification and supervised machine learning offers a reliable and objective alternative to manual scoring in neonatal HI models, enabling more efficient and reproducible behavioral assessment.

neuroscience↗