Biology subjects

Li, T.

Publications and source records attributed to Li, T..

At least 73 records · Page 4Linked to original sources

bioRxiv2026

Specification of bone marrow sinusoids requires TIE2-mediated positive feedback involving COUPTFII and VEGFR3

The bone marrow (BM) vascular network plays crucial roles in driving bone development and supporting hematopoiesis, yet the mechanisms governing its specialized architecture, particularly sinusoidal morphogenesis, remain inadequately characterized. We show in this study that TIE2 (Tek) was highly expressed by BM sinusoidal endothelial cells (SEC) and the endothelial Tek excision led to BM sinusoidal capillarization. Particularly, the BM sinusoids displayed thinner vessel diameter with the aberrant mural cell coverage in the Tek mutants. Mechanistically, TIE2 insufficiency led to a dramatic decrease of VEGFR3 in BM-SECs while its expression in hepatic sinusoids was not obviously altered. The RNA-seq analysis showed that GO terms enriched for the downregulated genes were related to the biological processes including sinusoidal development while pathways related to arterial ECs and angiogenesis were upregulated in the bone marrow of Tek mutants. The alteration of sinusoidal VEGFR3 expression occurred within 48 h after the induced endothelial deletion of Tek. Consistently, the defective BM sinusoidal formation was validated with the induced Tek deletion in VEGFR3+ SECs. The insufficiency of TIE2 ligand ANGPT1 also led to reduced sinusoidal VEGFR3, accompanied by similar BM sinusoidal defects. Furthermore, disruption of sinusoidal morphogenesis was observed in mutant mice with the endothelial excision of Nr2f2 (COUP-TFII), displaying a decreased expression of BM sinusoidal TIE2 and VEGFR3. These findings suggest that ANGPT1/TIE2 and COUP-TFII form a reciprocal regulatory loop to coordinate BM sinusoidal specification via regulating VEGFR3.

developmental biology↗

bioRxiv2026

Liver-to-Atria Inflammatory Axis Driving Arrhythmia

BackgroundMetabolic dysfunction-associated steatohepatitis (MASH) is emerging as a risk factor of cardiometabolic diseases, including the atrial fibrillation (AF) - the most common sustained arrhythmia. Given that the liver is a major source of inflammatory mediators, lipids, and hepatokines under metabolic stress, we hypothesized that hepatocyte-derived factors in MASH may accelerate atrial remodeling and arrhythmogenesis. MethodsAnalysis of the Atherosclerosis Risk in Communities (ARIC) visit 5 cohort was performed to determine the association between the FIB-4 index - a classic indicator of liver fibrosis, and AF risk, with multivariable adjustment for common comorbidities. A murine model of MASH was induced using the GAN (Gubra-Amylin NASH) diet. Programmed intracardiac stimulation and echocardiography were performed to assess AF susceptibility and cardiac function. Calcium imaging, histology, flow cytometry, plasma proteomics, and single-nucleus RNA sequencing (snRNA-seq) analyses were employed to elucidate the role of recruited inflammatory macrophages via hepatocyte-derived osteopontin (OPN) in MASH-induced atrial remodeling. ResultsAnalysis of the ARIC cohort confirmed a higher cumulative incidence of AF and an elevated adjusted hazard ratio (HR) in patients with intermediate and high FIB-4 indices compared to individuals with low FIB-4 scores. MASH mice exhibited increased susceptibility to pacing-induced AF, accompanied by enhanced proarrhythmic calcium release events, atrial enlargement, and fibrosis, independent of ventricular dysfunction. Proteomics and snRNA-seq revealed that the hepatocyte-secreted OPN under MASH conditions promoted the differentiation and recruitment of TGFBR1+ inflammatory macrophages to the atria, leading to gasdermin D (GSDMD) activation - an effector of inflammasome signaling and consequent proarrhythmic atrial remodeling. Activation of the monocyte-derived pro-inflammatory TGFBR1+ macrophages was dependent on the OPN receptor CD44. Furthermore, the MASH-induced atrial fibroinflammatory milieu and enhanced AF susceptibility were mitigated through several strategies, including hepatocyte-specific Spp1 (encoding OPN) deletion, neutralization of circulating OPN, ablation of CD44 or GSDMD. ConclusionsThese findings establish a pathogenic role of the hepatokine osteopontin in driving activation and recruitment of TGFBR1+ inflammatory macrophages into the atria, leading to proarrhythmic atrial remodeling under MASH. Osteopontin-targeted therapy or GSDMD inhibition prevents AF, indicating a novel therapeutic strategy for liver disease-related atrial arrhythmogenesis. Clinical PerspectiveO_ST_ABSWhat is new?C_ST_ABSO_LIIn the ARIC cohort, metabolic dysfunction-associated steatohepatitis (MASH) is associated with increased risk of atrial fibrillation (AF) after adjusting for common comorbidities. Elevated levels of circulating osteopontin (encoded by SPP1) predict an increased risk of AF in patients with MASH-induced liver fibrosis. C_LIO_LIMASH enhances hepatocyte secretion of osteopontin, leading to expansion of myeloid cells and recruitment of inflammatory macrophages into atria. This liver-to-atrial inflammatory circuit promotes the development of a substrate conducive to AF, which can be attenuated by hepatocyte-specific Spp1 deletion or neutralizing anti-anti-osteopontin antibody treatment to eliminate the mediator, or ablation of inflammasome effector gasdermin D to correct the atrial response. C_LI What are the clinical implications?O_LIOsteopontin may serve as a biomarker for AF in MASH cohorts. C_LIO_LIAnti-osteopontin therapy through neutralizing antibodies may serve as a novel therapeutic strategy for liver disease-related atrial arrhythmia. C_LI

systems biology↗

bioRxiv2026

Near-Infrared-II Chemo-Optogenetics for Deep-Brain Stimulation

Precise activation of ion channels enables fine-tuned control of neuronal excitability, providing a powerful strategy for dissecting and modulating neural circuits. Although optogenetics offers high spatiotemporal neuronal manipulation with cell-type specificity, its reliance on visible light (400-650 nm) limits tissue penetration to millimeter depths, restricting applications in deep-brain stimulation. Here we report a near-infrared-II (NIR-II)-sensitive calcium channel, HaloNeu, created by genetically fusing a circularly permuted HaloTag (cpHaloTag) to the thermo-sensitive transient receptor potential vanilloid 1 (TRPV1) and covalently conjugating NIR-II photothermal nanotransducers (HPN). The HaloNeu enables non-invasive neuromodulation at depths up to 1.0 cm at ultralow laser power ([~]60 mW/cm2) and up to 5.0 cm under the safe exposure limit ([~]1 W/cm2) with 1064 nm laser illumination. Remarkably, HaloNeu maintains stable, on-demand neuron-specific modulation for over two months in vivo, providing sustained activation of ventral tegmental area (VTA) circuits and effective alleviation of Parkinsonian symptoms in mouse models. These results establish HaloNeu as a robust and versatile platform for cell-type-specific, deep-tissue, and chronic neuromodulation, with broad implications for neuroscience and neurotherapeutics.

neuroscience↗

bioRxiv2026

Tauopathy primes co-filament assembly and dysfunction of TDP-43

While most Alzheimers disease (AD) which is associated with Limbic Predominant Age-related TDP-43 Encephalopathy (LATE) exhibits accelerated brain atrophy, the pathogenic mechanism remains elusive. We show here, in mice harboring depositions of amyloid-{beta} and tau, the age-dependent emergence of TDP-43 proteinopathy. We demonstrate that TDP-43 dysfunction facilitates caspase 3-mediated endoproteolysis of tau, accelerates tauopathy and exacerbates neuron loss. Unexpectedly, we found that the emergence and spread of TDP-43 proteinopathy is associated with the spread of tauopathy and correlated with co-filament assembly of tau and TDP-43. Importantly, TDP-43 dysfunction precedes such co-filament assembly and TDP-43 cytoplasmic aggregates. Consistent with the idea that tauopathy could prime co-filament assembly and proteinopathy of TDP-43 to exacerbate neurodegeneration, we found tau co-filament assembly with TDP-43 in AD and AD-LATE cases. These findings suggest that TDP-43 dysfunction accelerates tauopathy, which, in turn, primes co-filament assembly and dysfunction of TDP-43 to exacerbate neuron loss in AD-LATE, a pathogenic mechanism disclosing novel targets and therapeutic strategies.

neuroscience↗

bioRxiv2026

BiomniBench: Process-level Evaluation of LLM Agents for Real-world Biomedical Research

LLM agents now perform real biomedical research, but evaluating them rigorously is hard. Outcome-only benchmarks fail in two ways. First, a correct final answer can come from memorization, reward hacking, or wrong reasoning that produces the right number by chance. Second, valid alternative analyses are marked wrong simply because they differ from the reference. We introduce BiomniBench, a process-level evaluation framework that scores the full agent trajectory against expert-designed, task-specific rubrics. Our first release, BiomniBench-DA, contains 100 data-analysis tasks across 17 task types, 5 disease areas, and a general-biology category, each based on a paper from journals such as Nature, Cell, and Science and co-developed with an original author or a domain expert. Benchmarking frontier and open-weight models across four agent harnesses reveals three findings. Frontier and open-weight bases cluster within a few points of each other, with substantial headroom for all models. The agent harness shifts scores by more than the gap between successive model generations. Agents reliably ground claims in real sources yet consistently fall short on method selection, biological interpretation, and scientific reasoning. BiomniBench is the first process-level benchmark for LLM agents in biomedical research, providing the dimension-level diagnostics that outcome scoring cannot. Datasethuggingface.co/datasets/phylobio/BiomniBench-DA

bioinformatics↗

bioRxiv2026

Microglia-derived IL-18 remodels hippocampal plasticity to constrain traumatic fear memory

Post-traumatic stress disorder (PTSD) involves complex neuroimmune-synaptic crosstalk in the hippocampus. Here we show that interleukin-18 (IL-18), an extensively studied pro-inflammatory cytokine, serves a protective role against traumatic fear memory through a microglia-to-neuron signaling axis. Traumatic stress induces sustained upregulation of IL-18 in the hippocampus. Exogenous IL-18 administration attenuates fear memory, whereas blockade of IL-18 signaling exacerbates it. Mechanistically, microglial-derived IL-18 acts on neuronal IL-18R1 to restore stress-impaired synaptic plasticity and reduce perineuronal net density, thereby facilitating structural synaptic remodeling. In addition, IL-18 modulates the synaptic organization of fear memory-encoding engram cells within hippocampal ensembles. Together, these findings redefine IL-18 as a homeostatic regulator of post-trauma hippocampal synaptic function and identify the hippocampal IL-18 pathway as a potential therapeutic target for PTSD.

neuroscience↗

bioRxiv2026

geneSync: Gene Symbol Harmonization for Large-scale RNA-seq Data Integration

Cross-cohort integration of transcriptomic data is a routine strategy for boosting statistical power and enhancing generalizability. However, gene nomenclature inconsistencies across datasets--arising from annotation version updates, historical renaming, and synonym reassignment--introduce silent mismatches during feature alignment, causing genes to be falsely classified as absent or split into duplicate features. Here, we present geneSync, an R package that performs gene symbol harmonization as a quality-control (QC) step prior to data integration. geneSync uses a hierarchical matching strategy, prioritizing exact matches to authoritative gene symbols, then exact matches to National Center for Biotechnology Information (NCBI) gene symbols, and finally synonym-based fallback. It includes built-in offline databases for human, mouse, and rat, and supports auditable conflict resolution, cross-species ortholog mapping, and native integration with Seurat and SingleCellExperiment objects. Benchmarking across six mouse hippocampus scRNA-seq datasets spanning 2020-2025 and five CellRanger versions shows that 1.41%-6.22% of features require synonym resolution, and harmonization improves pairwise gene overlap by up to 13.14 percentage points, rescuing 707-1,098 genes per dataset pair. Notably, CellRanger annotation version--rather than data collection year--was identified as the primary driver of nomenclature discrepancy. geneSync is freely available at https://github.com/xiaoqqjun/geneSync.

bioinformatics↗

bioRxiv2026

Hippocampal Glucuronyl C5-epimerase promotes stress resilience by directly engaging PI3K through a non-enzymatic mechanism

Major depressive disorder (MDD) seriously affects human physical and mental health and causes global socioeconomic burdens. Stress-induced depressive etiology is linked to dendritic atrophy in the hippocampus, however, the underlying mechanism was poorly understood. Here, we identify that glucuronyl C5-epimerase (Glce), a heparan sulfae-modifying enzyme, as a critical regulator of hippocampal dendritic integrity and stress resilience. Genetic ablation of Glce in the hippocampal excitatory neurons is sufficient to induce dendritic atrophy and depressive-like behaviors, whereas its restoration rescues these deficits. Notably, Glce levels are reduced in the plasma of depressed individuals. We further find that Glce functions independently of its enzymatic activity, instead binding directly to and activating PI3K (p85/p110), thereby triggering an Akt/CREB/BDNF signaling cascade. Together, our findings uncover a non-canonical role for Glce and establish that Glce-PI3K-BDNF axis is essential for maintaining hippocampal structure and behavioral resilience, thereby highlighting a critical role of Glce in the pathophysiology of depression.

neuroscience↗

bioRxiv2026

Fibroblast Growth Factor Receptor Signaling in Maturing Osteoblasts Controls Cell-Matrix Interactions Critical for Osteocyte Survival

The terminal differentiation of osteoblasts into osteocytes, the most abundant cell type in cortical bone, is critical for skeletal homeostasis. Osteocyte loss is a hallmark of bone aging and fragility, yet the mechanisms regulating osteocyte formation and survival are poorly understood. We show that inactivation of fibroblast growth factor receptor 1 (Fgfr1) in the mature osteoblast lineage results in extensive osteocyte death, identifying FGFR1 signaling as essential for osteocyte viability and bone integrity. Lineage tracing and analysis of endogenous and induced appositional bone formation revealed that newly embedded osteocytes fail to survive without FGFR1. These osteocytes exhibited ectopic expression of osteocalcin and podoplanin within sclerostin-positive, TUNEL-reactive lacunae, along with defective dendrite formation and disruption of the local lacunocanalicular network. RNA sequencing of cortical bone demonstrated reduced expression of extracellular matrix (ECM) genes and neuronal regulatory genes, while histological and ultrastructural analyses showed disorganized collagen fibrils, diminished osteoid, and abnormal mineralization. In vitro, FGF signaling in Ocy454 cells regulated gene programs involved in development, axon guidance, and bone ECM organization, highlighting a dual function for FGF signaling in which it controls both matrix-dependent and intrinsic cell differentiation mechanisms during the osteoblast-to-osteocyte transition. We propose that FGFR1 deficiency causes ECM disorganization and impaired dendrite formation, disrupting osteocyte communication with neighboring bone and vascular cells, ultimately leading to cell death. These findings establish FGFR signaling as a critical regulator of osteocyte differentiation, viability of bone-embedded osteocytes, and bone homeostasis. Summary StatementFGFR signaling has a profound effect on adult bone extracellular matrix that is vital to maintaining the viability and morphology of newly formed osteocytes, their lacunocanalicular network and the maintenance of bone homeostasis.

developmental biology↗

bioRxiv2026

NbRD21 protease controls receptor kinase homeostasis in Nicotiana benthamiana

RD21-like proteases are papain-like cysteine proteases with a C-terminal granulin domain that are abundant and ubiquitous in angiosperms and have often been implicated in immunity. We previously found that the activity of RD21 in Nicotiana benthamiana (NbRD21) is suppressed during infection with Pseudomonas syringae. Here, we studied the role of NbRD21 in immunity and proteome processing. NbRD21 was disrupted by genome editing and rd21 mutants were subjected to disease assays and shot-gun proteomics. Dipeptide substrate zLR-AMC was used in protease assays and agroinfiltration was used to transiently express NbRD21 and candidate substrates. Genome edited lines lacking NbRD21 develop normally but have drastically reduced zLRase activity and are significantly more susceptible to P. syringae. Shot gun proteomics revealed an increased accumulation of [~]20 diverse receptor-like kinases (RLKs) in untreated rd21 knockout lines, but their transcript levels are unaltered when compared to wild-type plants. 35S-driven GFP-tagged RLKs accumulate more upon transient expression in rd21 plants than in wild- type plants. These data indicate that NbRD21 post-translationally controls RLK homeostasis, either by directly degrading RLKs, or indirectly by regulating endocytic RLK recycling.

plant biology↗

bioRxiv2026

Multi-scale dissection, compaction and derivatization of mammalian developmental enhancers

Gene expression during mammalian development is orchestrated by non-coding cis-regulatory DNA elements (CREs) such as distal enhancers1-3. Despite their fundamental importance, and notwithstanding recent progress in predictive modeling4-9, many high-level properties of enhancer grammar remain unresolved. How does the length of an autonomously active CRE constrain its activity? How robust are CREs to mutations or rearrangements of transcription factor binding sites (TFBSs)? And how much epistasis exists among these sites? As predictive models solely trained on endogenous CREs are unlikely to resolve these questions10, we subjected several endogenous CREs to intensive sequence-level perturbation. Specifically, we assayed >35,000 variants of 5 parietal endoderm enhancers, with variants organized into four perturbation classes, designed to probe: (i) the functional sufficiency of sub-fragments via dense multi-size tiling, (ii) local epistasis via multi-hit saturation mutagenesis, (iii) activity-size tradeoffs via model-guided compaction, or (iv) functional resilience via sequence derivatization anchored on key TFBSs, including random deposition, reconstitution, and synthetic thripsis. This multi-scale dissection revealed rich phenomena. Sub-tiling uncovered sharp non-additivity between activity and fragment size, highlighting strongly synergistic TFBS clusters. Compaction showed that natural CREs lie far from the activity-size Pareto front, and that model-guided deletions can yield shorter yet stronger elements. Mutational scanning exposed a spectrum of CRE robustness, from tolerant to fragile, together with rare but consequential epistasis between individual TFBSs. Finally, TFBS-anchored derivatization demonstrated that background sequence can influence activity on par with TFBS arrangement. Strikingly, a substantial fraction of CRE derivatives exceeded the activity of their endogenous progenitors. Taken together, these results reveal both soft and stiff directions in regulatory sequence space, advancing a quantitative phenomenology of how enhancer sequences encode function and robustness.

genomics↗

bioRxiv2026

Retracing and rewriting the evolutionary trajectories of mammalian developmental enhancers

Cis-regulatory elements (CREs) such as enhancers play a central role in orchestrating mammalian development, yet how they have gained, lost, maintained or changed function over the course of mammalian evolution remains poorly understood. To address this gap, we densely mapped the functional evolution of five mouse developmental enhancers by testing orthologous sequences from 480 extant and ancestrally reconstructed mammalian genomes (Zoonomia1, Cactus2) with massively parallel reporter assays (MPRAs). This phylogenetic dissection revealed diverse modes of evolution, from lineage-restricted activity to deep functional conservation despite extensive sequence divergence. To pinpoint causal changes, we developed a model-driven reconstitution strategy that uses deep learning-based predictions of chromatin accessibility to re-introduce a succession of mutations into ancestral orthologs; this revealed critical transcription factor binding site (TFBS) changes and pervasive context-dependent epistasis, including instances where mutational effects were strongly contingent on the order of their introduction. When we extended this strategy to tune the activity of extant orthologs, we found that ablation of enhancer function required as few as one to seven mutations, whereas enhancement was constrained by element-specific activity ceilings--a striking asymmetry in the predictability of model-guided enhancer editing. Together, these results shed light on how the plasticity of mammalian enhancers intersects with their evolution, and advance a framework for reprogramming the activity of endogenous CREs at nucleotide resolution.

molecular biology↗

bioRxiv2026

Cardiac oxidative stress monitoring enabled by hierarchical mechanical adaptation

Soft bioelectronics have advanced cardiac monitoring through electrophysiological tracking, yet this alone cannot resolve the metabolic pathology essential to surgical decision-making. However, real-time molecular sensing on beating hearts remains unresolved due to deformation-induced sensor failure and stress-induced metabolite artifacts. This challenge is exemplified by ischemia-reperfusion injury (IRI), a major cardiac surgery complication characterized by reactive oxidative species (ROS) bursts, where true pathological ROS signals being confounded by mechanotransduction-induced ROS artifacts. Herein, we propose an enzymatic cardiac oxidative stress biosensor (E-cardiac) with hierarchical mechanical adaptation: macro-scale biofluid-mediated contact, micro-scale fiber reorganization, and nano-scale enzymatic confinement within gold nanoarches dissipate interfacial stress. This produces ultrathin ([~]460 nm), soft (0.79 kPa) E-cardiac with robust electrochemical stability (100% strain), low detection limit (380 nM), rapid adhesion (<3 s), stable biosensing on beating heart, as well as minimal invasive deployment capability. Mechanical analysis and cellular studies confirm mitigated stress-induced ROS and absent PIEZO channel activation. Validated across cardiomyocytes, ex vivo tissues, multi-species ischemia models (mouse, rat, rabbit, pig), rat ischemia-reperfusion injury, and Langendorff hearts simulating graded perfusion deficits, E-cardiac quantitatively differentiates IRI severity (sham < ischemia < reperfusion) as well as detecting the "ECG blind window". The E-cardiac platform provides real-time metabolic feedback for surgical guidance during cardiac procedures, enabling timely intervention before irreversible damage.

bioengineering↗

bioRxiv2026

Multiscale Mechanisms of Human Memory Modulation by Deep Brain Stimulation

The highly variable effects of deep brain stimulation (DBS) on cognition highlight a fundamental gap in our understanding of how external perturbations interface with the brains ongoing, task-engaged state. The current study addressed this challenge by combining intracranial EEG recordings with hippocampal DBS stimulation in human participants during spatial sequence memory task. We found 50 Hz stimulation targeting hippocampal white matter enhanced memory performance, whereas 5 Hz stimulation targeting hippocampal gray matter impaired memory. These opposing behavioral outcomes were both governed by two dissociated mechanisms: a region-specific modulation of theta rhythms in an engagement-dependent fashion, and a global modulation of the fidelity of memory representations across the cortex. Our results suggest targeted stimulation can concurrently fine-tune specialized circuits and govern the brains global state, which not only offers a novel roadmap for understanding how external inputs shape human cognition, but also helps to design precise, state-dependent neuromodulation therapies.

neuroscience↗

bioRxiv2026

Connectome-based spatial statistics enabling large-scale population analyses of human connectome across cohorts

Large-scale population analyses of structural connectome organization remain challenging because of cross-subject alignment, pathway interpretability and computational burden. No widely adopted standard exists for systematic evaluation across processing methods. We developed connectome-based spatial statistics (CBSS), a scalable framework for anatomically aligned and functionally informed quantification of white-matter microstructure that yields atlas-defined pathways organized into 13 functional networks. Using data from 56,510 UK Biobank participants together with five independent lifespan cohorts, we evaluated the streamline-, voxel- and network-level measures in the aspects of reliability, heritability, structure-function coupling, cognitive and behavioral prediction, brain aging patterns and lifespan trajectories across cohorts. The systematic evaluation workflow compares population-level white-matter representations across methods, spatial scales, tasks and datasets. The results support CBSS as a common connectome reference for large-scale, cross-cohort diffusion MRI studies.

neuroscience↗

bioRxiv2026

Curvature-mediated prewetting organize mitochondrial nucleoid

The emergence of spatiotemporal order in diffusive cellular environments requires physical mechanisms to overcome the entropic drive toward disorder. While the organizing role of chemical signaling is well characterized, how complex membrane curvature regulate soluble molecular organization remains a fundamental open question. Here, we establish membrane curvature as a thermodynamic control parameter for the condensation of soluble proteins through prewetting. Using the mitochondrial transcription factor TFAM, we demonstrate that intrinsic membrane curvature locally drives a prewetting transitiona surface-mediated phase separation distinct from bulk condensation. By combining in vitro reconstitution, live-cell super-resolution imaging, cryo-electron tomography, and thermodynamic theory, we show that negatively high-curvature regions lower the nucleation energy barrier of Tfam, driving localized protein condensation at physiological concentrations. These results reveal that membrane curvature does not merely scaffold cellular machinery but manipulates the local free energy landscape, establishing a geometric principle for the spatial control of biological processes.

biophysics↗

bioRxiv2026

Evolutionary transfer learning enables organism-wide inference of mammalian enhancer landscapes

Understanding and modeling how a single human genome concurrently encodes gene regulatory programs for thousands of cell types remains a central challenge in genomics and machine learning. Most human cell types emerge during embryonic, fetal, and pediatric development which are inaccessible to comprehensive molecular profiling. To circumvent this, we hypothesized that the mismatch in evolutionary rates between cis-acting enhancers that modulate gene expression (fast) and the trans-acting regulatory factors that specify cell types (slow) creates an opportunity for evolutionary transfer learning. Specifically, models trained to predict cell type-specific enhancers in one species should generalize to the orthologous cell types and enhancers of related species. To test this, we generated a single-cell atlas of chromatin accessibility spanning mouse embryonic day 10 (E10) to birth (P0). Using combinatorial indexing1, we profiled 3.9 million nuclei from 36 staged embryos, resolving genome-wide accessibility in 36 cell classes and 140 cell types. We then trained a series of multi-output deep learning models (CREsted2), each addressing limitations of the preceding approach, towards the goal of genome-wide prediction of distal enhancers across major developmental lineages. An evolution-naive model achieved strong performance on heldout peaks, but exhibited two failure modes during genome-wide inference: overprediction at tandem repeats and conflation of promoter and distal enhancer grammars. An evolution-aware model resolved these by regrouping accessible regions based on their retention and functional coherence across mammalian evolution, but failed to generalize across species. Finally, an evolution-augmented model, STEAM (Synteny-aware Transfer learning for Enhancer Activity Modeling), incorporated enhancer orthologs from 241 mammalian genomes (Zoonomia3) in a synteny-supervised manner. This increased the effective data scale by as much as 195-fold, markedly improving generalization across mammals despite greater label noise. We applied STEAM to the genome-wide inference of cell class-specific distal developmental enhancers in humans, mice (HumMus) and 239 additional mammals3 (BabaGanoush), i.e. 32 x 241 = 7,712 genome-wide distal enhancer prediction tracks. Together, our results unify advances in single-cell profiling, deep learning, and comparative genomics into a framework for the evolutionary transfer learning of noncoding regulatory grammars. More broadly, our work supports the view that model organisms and evolutionarily diverse genomes are indispensable resources for accelerating and enhancing the AI-enabled exploration of human biology. NoteAn interactive version of this preprint, together with count matrices, CREsted models, prediction tracks, code and reproducible figures, is available at this link (ref 4).

genomics↗

bioRxiv2026

In vivo validation of predicted fitness effects at single-base resolution in a Brachypodium distachyon mutant population

Computational tools, including biological language models (LMs), show substantial promise in predicting the impact of genetic variants on plant fitness. However, validating variant effect predictions (VEP) requires experimental populations where genetic variation consists of discrete point mutations rather than segregating recombination blocks. In this study, we generated a novel population of Brachypodium distachyon mutant lines to evaluate the accuracy of VEP at single-base resolution. These lines were advanced through single-seed descent for five generations (M1 to M5), with whole-genome sequencing performed at M2 and M5 and phenotypic measurements recorded at M3 and M4. Using state-of-the-art VEP models, we predicted the functional impact of missense protein-coding variants and gene-proximal non-coding variants. We validated these predictions by estimating the effect of mutations on whole-plant measurements (burden tests) and their probability of fixation from M2 to M5 (purging tests). Among missense variants, the protein LM ESM showed superior predictive accuracy compared to the bioinformatic standard SIFT and the genomic LM PlantCAD. Notably, the relationship between VEP scores and allele fixation suggested a log-linear relationship between VEP scores and variant fitness. Among gene-proximal variants, PlantCAD appeared more accurate than supervised models of regulatory activity, such as chromatin accessibility (a2z) and RNA abundance (PhytoExpr). Collectively, our findings highlight the utility of state-of-the-art VEP tools as predictors of fitness and demonstrate the potential of mutant populations to evaluate computational tools for precision breeding applications.

genomics↗