Biology subjects

Li, H.

Publications and source records attributed to Li, H..

At least 343 records · Page 19Linked to original sources

bioRxiv2025

Regulatory logic of neuronal identity specification in Drosophila

Combinations of terminal selector transcription factors (tsTFs) are thought to establish and maintain the unique identities of the numerous cell types found in nervous systems. However, it remains largely unclear how tsTF combinations are specified during development, and how they then coordinate the type-specific differentiation programs of each neuron. To investigate these regulatory mechanisms, we performed simultaneous single-cell RNA and ATAC sequencing on the Drosophila optic lobes at four stages of their development and identified over 250 distinct cell types. We characterized the common cis-regulatory features of neuronal enhancers and performed comprehensive inference of gene regulatory networks across cell types and stages. Our results reveal cell-type and stage-specific enhancers of many neuronal genes and the cooperative actions of tsTFs, pan-neuronal and ecdysone-responsive TFs on these enhancers. We show that the same effector genes are often regulated by different tsTF combinations acting through distinct enhancers in different neurons. During neurogenesis, tsTF codes are established within a brief critical period in newborn neurons, often through cell-type-specific enhancers that are not accessible in their progenitors. Accordingly, when neuroblast temporal patterning TFs are re-utilized as tsTFs in neurons, they are regulated independently through separate enhancers. Therefore, neuronal identity specification and differentiation is a multi-step regulatory program, wherein the same TFs enact distinct regulatory codes at different steps and across cell types.

neuroscience↗

bioRxiv2025

Combined multi-color immunofluorescence staining and spatial in situ mRNA expression analysis identifies potential fibrosis drivers in acute lymphoblastic leukemia

Acute lymphoblastic leukemia (ALL) is the most prevalent childhood cancer. Bone marrow (BM) fibrosis in ALL has been associated with adverse outcomes, however, little is known about the mechanisms that cause fibrosis in ALL. Therefore, we established a novel and advanced analysis method by combining multi-color immunofluorescence staining with in-situ RNA expression analysis (RNAscope(R)) investigate the spatial expression of putative fibrotic drivers in ALL bone marrows. We analyzed standard BM biopsies from pediatric ALL patients. Sequential 5-color immunofluorescence (IF) staining with CD45, CD271, CD31, CD34 and DAPI was used to identify different BM cell types. Combined RNAscope(R) and IF staining was established for spatial mRNA expression analysis of transforming growth factor beta 1 (TGFB1) and platelet-derived growth factor alpha 1 (PDGFA1), which are known to play major roles in primary myelofibrosis (PMF). PMF and normal BM samples served as controls. As expected, ALL bone marrows showed high cellularities and prominent populations of blast cells. CD271+ MSC density was increased in ALL and was associated with fibrosis in a similar manner as observed for PMF. TGFB1 and PDGFA1 expression was considerably increased in ALL megakaryocytes (MKs) compared to PMF patients and normal controls. Furthermore, MK TGFB1 and PDGFA1 expression intensities in fibrotic ALL correlated with fibrosis grade. TGFB1 and PDGFA1 were also expressed in leukemic blasts, however at lower intensities compared to ALL MKs. Taken together, advanced in-situ RNA and IF staining not only revealed increased expression of TGFB1 and PDGFA1 in fibrotic pediatric ALL, but also identified ALL blasts and MKs as their cellular origin at the single cell level. These novel data strongly suggest a role of these cytokines as potential fibrosis drivers in ALL. More broadly, our findings demonstrate that combined RNA and surface marker analysis is a powerful tool to provide new and valuable insights into bone marrow pathophysiology.

cell biology↗

bioRxiv2025

Antitachycardia Pacing is More Effective When Delivered to the Left Bundle Branch Area Compared to the Right Ventricle in a Pre-clinical Model of Ischemic Ventricular Tachycardia

BackgroundAnti-tachycardia pacing (ATP) delivered from implantable cardioverter defibrillators (ICDs) provides critically timed pacing pulses to terminate ventricular tachycardia (VT). Physiological pacing through left bundle branch area (LBBA) pacing has emerged as a clinically relevant alternative to induce synchronous activation of the ventricles. The main objective of this study was to compare the efficacy and safety of ATP delivered to an LBBA lead and a conventional RV lead. MethodsUsing a preclinical animal model (n=7), pacing leads were implanted in the RV apex and LBBA and connected to ICDs. The left anterior descending artery was occluded for two hours to cause an ischemia-reperfusion injury (IRI). Four days following IRI, VT episodes were induced using programmed electrical stimulation (PES), and burst ATP therapy was delivered to the RV or LBBA leads for each VT episode. ResultsVT was induced 80 times, with a mean VT cycle length (VT CL) of 180.0{+/-}30.0 ms. ATP delivered to the LBBA terminated VT more often than RV ATP (70.2% vs 47.3%, P = 0.04), and there was no significant difference in negative outcomes of VT acceleration or VF induction. Activation sequences from the basket catheter were determined for the pre-therapy VT and captured ATP beats, and correlation coefficients were computed to the activation sequences during ATP delivery. The number of ATP pulses required for the activation sequences to correlate with the captured ATP pattern rather than the pre-therapy VT pattern was lower for LBBA ATP compared to RV ATP (4.1 vs. 5.1 beats, P = 0.04). Successful LBBA ATP demonstrated earlier and more frequent Purkinje activations preceding myocardial activation than successful RV ATP. ConclusionImproved performance of LBBA ATP compared to RV ATP provides further incentive for the wide use of LBBA leads in patients who need cardiac electrotherapy.

bioengineering↗

bioRxiv2025

The multiscale distribution of radiation-induced DNA damage and its impact on local genome structure

The three-dimensional genome structure is critical for the regulation of gene expression and repair of DNA damage. While previous work has characterized genome-wide sites of DNA damage caused by etoposide or nucleases, the distribution of double-strand breaks (DSBs) caused by external radiation and how these interact with the 3D genome organization is less well understood. Here, we measure the genomic landscape of radiation-induced DNA damage using END-seq in fibroblasts and lymphoblasts after exposure to 5 Gy X-rays. We identify frequently broken regions and investigate the 3D genome properties around these breaks with Hi-C data. We observe that the distribution of robust breaks correlates with transcriptional and chromatin features of the genome. Transcriptionally active and decondensed regions, such as chromosome 19, the A compartment, and topologically associating domain (TAD) boundaries, show pronounced break probability. We also find evidence of DSB-induced loop formation in the vicinity of frequent radiation-induced breaks. Our data reveal that pre-existing 3D genome architecture influences the distribution of radiation-induced DSBs and that these breaks reshape local chromatin landscapes.

genomics↗

bioRxiv2025

Stomatin encapsulates aquaporin-1 and urea transporter-B in the erythrocyte membrane

Stomatin is a ubiquitous and highly expressed protein in erythrocytes, which associates with cholesterol-rich microdomains in the plasma membrane and is known to regulate the activity of multiple ion channels and transporters, but the structural basis of association with stomatin targets remains unknown. Here we describe high-resolution structures of multiple stomatin complexes with endogenous binding partners isolated from human erythrocyte membranes, revealing that stomatin specifically associates with two membrane proteins involved in water transport and cell volume regulation, aquaporin-1 (AQP-1) and the urea transporter, UT-B (SLC14A1). Together, our results reveal the structural basis of stomatin oligomerization, membrane association, and target recruitment, and identify a putative role for stomatin in the regulation of osmotic balance in the erythrocyte.

biochemistry↗

bioRxiv2025

pH-driven evolutionary divergence and ongoing speciation in the cosmopolitan cyanobacterium Microcoleus vaginatus

Speciation models and the concept of speciation continuum have been well established in multicellular eukaryotes. However, the mechanisms underlying population differentiation and their environmental drivers in prokaryotes remain elusive. Here, we explored the population diversification of the filamentous cyanobacterium Microcoleus vaginatus, a photoautotrophic edificator with significant importance for dryland ecosystems worldwide. By constructing a 132-genome dataset spanning multiple climate zones across four continents, we integrated intraspecific genomics with environmental association analyses to elucidate the population genetic structure and molecular mechanisms underlying the divergence of the M. vaginatus continuum. Our findings reveal that adaptive differentiation is driven by the combined effects of natural selection, homologous recombination, and gene family turnover, with varying contributions across different phylogroups. The evolution of core and accessory genomes exhibits complementary patterns, with genome-wide association studies demonstrating the predominant role of the core genome in population divergence. Selection, followed by neutral evolution and recombination, mediates trade-offs in life-history strategies, while environmental factors shape these evolutionary processes. Notably, while local edaphic variables exert direct influences on Tajimas D values, they demonstrate predominantly indirect modulation through alterations in the selection-to-gene gain ratio (S/G). pH emerges as the principal abiotic driver governing population genetic signatures, operating through coordinated effects on both S/G and the proportion of selected genes undergoing recombination (SUR). This study offers novel insights into the diversification mechanisms of cosmopolitan, dominant cyanobacteria, thereby advancing our understanding of prokaryotic speciation in complex natural settings and providing a framework for predicting microbial adaptation in a changing world.

evolutionary biology↗

bioRxiv2025

STARNet enables spatially resolved inference of gene regulatory networks from spatial multi-omics data

Biological tissues are composed of distinct microenvironments that spatially orchestrate gene expression and cell identity. However, the regulatory principles governing domain-specific cellular functions remain poorly understood due to the lack of effective methods for mapping gene regulatory networks (GRNs) in situ. To address this gap, we introduce STARNet, a representation learning approach that leverages heterogeneous hypergraph modeling of spatial transcriptomic and epigenomic data to resolve tissue-domain-specific regulatory interactions. By integrating graph neural networks with contrastive learning in a self-supervised framework, STARNet learns unified embeddings that preserve both multi-modal molecular features and anatomical spatial context, enabling accurate and domain-resolved GRN reconstruction within complex tissues. Benchmarking on both simulated and real datasets demonstrates that STARNet achieves state-of-the-art performance. We further demonstrate its broad applicability across diverse biological contexts, including neural development, genetic disease risk, and drug-induced developmental toxicity. In the mouse brain, it delineates region-specific regulatory networks and reconstructs spatiotemporal programs underlying neural stem cell differentiation. In human genetics, it provides a mechanistic link between genotypes and phenotypes by showing how genome-wide association study (GWAS) variants for complex diseases perturb hippocampus-specific GRNs. In developmental toxicology, STARNet reveals that drug-induced disruptions of GRNs in defined embryonic regions underlie tissue-specific vulnerability. Collectively, STARNet offers a powerful and versatile framework for resolving the spatial regulatory logic of complex tissues, providing multi-angle insights into tissue patterning, development, and disease mechanisms.

bioinformatics↗

bioRxiv2025

An integrated multi-tissue atlas of epigenomic landscapes and regulatory elements in the bovine genome

Deciphering the regulatory syntax of the genome is essential to understand the genetic and molecular architecture of complex traits, as most trait-associated variants lie in non-coding regions. Yet, functional annotation of the bovine genome remains limited, hindering our ability to unravel the mechanisms underpinning complex traits of economic and ecological importance in cattle. Here, we present a comprehensive epigenetic atlas comprising 1,138 genome-wide epigenetic profiles, including chromatin accessibility, six histone modifications, CCCTC-binding factor (CTCF) transcription factor binding, DNA methylation, chromatin conformation, and transcriptomes across 53 adult tissues, five fetal tissues, and seven primary cell types. This atlas-level data enables us to annotate around 45% of the genome as putative regulatory elements exhibiting tissue- or cell-specific regulatory activity. Leveraging sequence-to-function deep learning models, we discovered 301 sequence motifs and predicted the functional impact of genetic variants through in silico mutagenesis, thereby facilitating the decoding of the regulatory syntax of the cattle genome and fine-mapping of GWAS loci for 22 complex traits. Cross-species analysis further revealed evolutionarily conserved features of regulatory architecture and provided evolutionary insights into complex traits and diseases in humans. Together, this atlas offers a foundational resource for advancing cattle functional genomics, sustainable breeding, and studies of regulatory evolution.

genomics↗

bioRxiv2025

High-throughput 3D super-resolution ultrasound imaging

Capturing fast hemodynamics in deep organs is essential for understanding microvascular regulation of organ responses to physiological demands and pathological stress. However, non-invasive three-dimensional (3D) imaging of these microscale processes remains challenging due to trade-offs between spatial resolution, imaging speed, and penetration depth. Here, we present fluctuation-based high-order super-resolution acoustic microscope (FLAME), a tracking-free 3D ultrasound imaging technology capable of fast microvascular angiography and flow measurement. Using as few as 30 volumes, FLAME improves 3D ultrasound resolution by 8-fold ([~]50 m) and shortens data acquisition time by over two orders of magnitude, from tens of seconds to tens of milliseconds, compared with conventional tracking-based super-resolution approaches1-6. This high-throughput capability supports high-fidelity imaging of tissue functions under flexible experimental settings. FLAME achieves an unprecedented 3D super-resolution frame rate of [~]40 Hz, capturing transient hemodynamic responses to various physiological and pathological stimuli in different mouse organs, from the microvascular level to the whole-body scale. With open-sourced implementation, FLAME provides an accessible platform for real-time, in-depth hemodynamic monitoring across diverse biomedical applications and clinical translation.

bioengineering↗

bioRxiv2025

Pyrimethamine inhibits mutant NRF2 as a molecular glue

Esophageal squamous cell carcinoma (ESCC) is a deadly disease and one of the most aggressive cancers of the gastrointestinal tract. As a master transcription factor regulating the stress response, nuclear factor erythroid 2-related factor 2 (NRF2) is often mutated, loses negative regulation by Kelch-like ECH-associated protein 1 (KEAP1), becomes hyperactive, and thus causes chemo-radioresistance and poor survival in human ESCC. Our previous research identified pyrimethamine (PYR) as an NRF2 inhibitor and demonstrated that inhibition of dihydrofolate reductase (DHFR) is the major mechanism of its action. Based on these findings, PYR has advanced into a Phase I window-of-opportunity clinical trial. In this study, using isogenic ESCC cell lines, we aimed to investigate the phenotypic consequences of NRF2 activation in ESCC cells and further elucidate the mechanism of action of PYR. While overexpression of wild-type NRF2 (NRF2WT) promoted squamous differentiation and suppressed proliferation, expression of mutant NRF2 (NRF2Mut) produced mixed effects on differentiation markers and promoted proliferation. NRF2 activation reduced sensitivity to chemotherapy and radiation, whereas PYR treatment restored chemosensitivity and radiosensitivity in NRF2Mut ESCC cells. Co-immunoprecipitation and proximity ligation assays revealed that PYR selectively enhanced the interaction between NRF2W24C and KEAP1, but not between NRF2WT and KEAP1. Isothermal titration calorimetry (ITC) confirmed the direct binding of PYR to KEAP1 and surface plasmon resonance (SPR) showed that PYR modestly facilitated the interaction between a DLGW24C peptide and the Kelch fragment. Molecular docking suggested that PYR bound to a pocket within the Kelch domain near Arg415. In summary, NRF2 activation promotes cell proliferation and therapy resistance in ESCC in a context-dependent manner. PYR functions as a modest molecular glue that selectively restores KEAP1 binding to NRF2W24C, providing a potential therapeutic strategy for NRF2Mut ESCC.

cancer biology↗

bioRxiv2025

Loss of brain insulin production impairs learning and memory in female mice

Diabetes is characterized by dysfunctional insulin release and action, and it is a risk factor for Alzheimers disease, the most common form of dementia. Alterations in brain insulin signalling and metabolism have been linked with Alzheimers disease. The ancestral insulin gene, Ins2, is transcribed locally within the brain. Here we demonstrate that hippocampal Ins2 mRNA is higher in females than males, and modulated by diet. To specifically determine how locally-produced insulin influences hippocampal function, we used mice with germline Ins2 knockout (Ins2-/-) and the normal complement of wildtype Ins1 alleles. Compensation from the Ins1 gene ensured normal glucose tolerance, normal insulin sensitivity, normal fasting insulin, and normal body weight under these conditions. Using the Morris water maze, we found that visuo-spatial learning and memory performance of female Ins2-/- mice was significantly impaired relative to wild-type mice, whereas the performance of male Ins2-/- and wild-type mice did not differ. We profiled isolated hippocampi from female Ins2-/- and littermate control mice using RNA sequencing to provide an unbiased analysis of gene expression differences that underlie these behavioural changes. Cyclin D1 (Ccnd1) was significantly reduced in Ins2-/- mice, prompting us to examine adult neurogenesis using exogenous mitotic marker EdU and the immature neuronal marker doublecortin (DCX). Our data demonstrate female-specific roles for brain-derived Ins2 on learning and memory function in mice. Research in contextO_ST_ABSWhat is already known about this subject?C_ST_ABSO_LIAlzheimers disease and type 2 diabetes share common risk factors and metabolic dysfunction. C_LIO_LIInsulin and insulin signalling are known to have positive effects on neuronal health and cognition. C_LIO_LILocal insulin production has been reported in the brain, including the hippocampus, which plays critical roles in learning and memory, as well as Alzheimers pathophysiology. C_LI What is the key question?O_LIWhat are the consequences of deleting the Ins2 gene, which selectively prevents local brain insulin production, on learning and memory in mice? C_LI What are the new findings?O_LIHippocampal insulin production via the Ins2 gene is sexually dimorphic and modulated by diet. C_LIO_LIKnocking out brain insulin production in Ins2-/- mice does not have significant effects on glucose homeostasis in these diet and housing conditions. C_LIO_LIKnocking out brain insulin production in Ins2-/- mice results in learning and memory impairments in female but not male mice. C_LI How might this impact on clinical practice in the foreseeable future?O_LIThese pre-clinical observations demonstrate a causal role for insulin in learning and memory, but will require additional human studies before they can impact clinical practice. C_LI

physiology↗

bioRxiv2025

Three-dimensional spatial transcriptomics at isotropic resolution enabled by generative deep learning

Mapping the complete three-dimensional (3D), transcriptome-wide spatial architecture of tissues and organs remains a fundamental challenge in biology. Typically, studies approximate 3D structures by profiling serial two-dimensional (2D) tissue sections, resulting in incomplete and anisotropic reconstructions. Here, we demonstrate that organ-wide, isotropic-resolution 3D gene expression landscapes can be effectively inferred from sparse 2D spatial transcriptomics data using artificial intelligence. Our framework, isoST, leverages stochastic differential equations (SDEs) to model the intrinsic biological continuity of gene expression along tissue depth, enabling the integration of discrete 2D measurements into coherent 3D transcriptomic maps. isoST is applied across diverse biological systems, including the brain, kidney, spinal cord, and gastrulating embryos, accurately reconstructing anatomical structures and molecular patterns that are only partially resolved in 2D analyses. isoST enables the comprehensive characterization of tissue niches at full 3D resolution, uncovering radial and spherical gene expression gradients associated with disease progression that are often distorted or missed in 2D approaches. Furthermore, isoST is extended to integrate histological imaging for reducing experimental complexity and to reconstruct high-resolution developmental spatial trajectories over time. Together, isoST provides a practical and scalable solution toward comprehensive high-resolution 3D spatial transcriptomic atlases.

bioinformatics↗

bioRxiv2025

Helicobacter pylori genome aggregation database reveals complex evolutionary forces shaping its genomic landscape and clinical impact in East Asia

We assembled HPgnomAD, a comprehensive global genome aggregation database of Helicobacter pylori (H. pylori) featuring 7,544 high-quality genomes from 278 populations, along with the first species-wide haplotype reference panel for genotype imputation. The panel provides high accuracy across diverse H. pylori populations, including those from low-coverage data, and is openly accessible with integrated analysis tools (https://www.hpgnomad.top/). Variant discovery revealed 1.82 million SNPs and 0.65 million InDels, with African strains exhibiting the most tremendous diversity and East Asian strains showing unexpectedly high novel variation. Fine-scale analysis of the hpEAsia lineage revealed six new sublineages shaped by altitude-, latitude-related divergence and region-specific admixture. Phylogenomic dating revealed two divergence waves in East Asia, paralleling Upper Paleolithic settlement and Neolithic human expansions, with highland-lowland separation at [~]18.8 kya, followed by the formation of complex geography-related population substructures. Genome-wide scans revealed adaptive loci related to metal acquisition, nitrogen metabolism, surface adhesion, and membrane transport, including altitude-associated highly differentiated variants linked to antibiotic resistance and gastric disease. By integrating genomic, evolutionary, environmental, and clinical data, HPgnomAD offers a framework for understanding H. pylori evolution, host-pathogen adaptation, and precision medicine worldwide.

evolutionary biology↗

bioRxiv2025

Mutant IDH silences GSX2 to reprogram neural progenitor cell fate and promote gliomagenesis

Isocitrate dehydrogenase (IDH) mutations arise early in gliomas and are associated with a defined neurodevelopmental cancer cell hierarchy. However, how mutant IDH contributes to this hierarchy and whether this interaction promotes gliomagenesis remain unclear. We captured the dynamics of IDH-mutant glioma initiation in genetically engineered mice through time-resolved, single-cell genomics. Mutant IDH activates and induces lineage switching of neural progenitor cells (NPCs). These actions expand oligodendrocyte precursor cells, the predominant cell-of-origin for these tumors, at the expense of interneurons. Lineage switching is mediated by promoter hypermethylation and silencing of Gsx2, a homeobox gene required for neurogenesis. Critically, Gsx2 ablation recapitulates NPC fate reprogramming by mutant IDH. We provide a new model of neural cell fate control by IDH oncogenes and insights into the developmental origins of glioma.

cancer biology↗

bioRxiv2025

A prevalent huge phage clade in human and animal gut microbiomes

Huge phages are widespread in the biosphere, yet their prevalence and ecology in the human gut remain poorly characterized. Here, we report Jug (Jumbo gut) phages with genomes of 360-402 kilobase pairs that comprise [~]1.1% of the reads in human gut metagenomes, and are predicted to infect Bacteroides and/or Phocaeicola. Although three of the four major groups of Jug phages shared >90% genome-wide sequence identity, their large terminase subunits exhibited only 38- 57% identity, suggesting horizontal acquisition from other phages. Over 1,500 genomes of Jug phages were recovered from human and animal gut metagenomes, revealing their broad distribution, with largely shared gene content suggestive of frequent cross-animal-host transmission. Jug phages displayed high gene transcription activities, including the gene for a calcium-translocating P-type ATPase not detected previously in phages. These findings broaden our understanding of huge phages and highlight Jug phages as potential major players in gut microbiome ecology.

microbiology↗

bioRxiv2025

UCP1 Mitigates Hepatic Steatosis and Fibrosis Independent of Cold Exposure

Non-shivering thermogenesis by brown adipose tissue (BAT) is a promising target for anti-obesity therapies, making its regulatory mechanisms of significant translational interest. While cold-induced BAT thermogenesis (CIT) is well characterized, certain high-calorie diets can also activate BAT in the absence of cold, a process known as diet-induced thermogenesis (DIT). Despite its potential relevance to modern human diets and lifestyles, the mechanisms and physiological relevance underlying DIT remain poorly understood. Here, we show that DIT reduces adiposity and protects against hepatic steatosis and fibrosis in males but not female mice. Moreover, adipose tissue-specific ablation of uncoupling protein 1 (UCP1) reveals that BAT is the primary mediator of DIT but that non-adipocyte UCP1 also contributes to body weight regulation. Transcriptome analysis suggests that DIT is triggered by intrinsic metabolic stress, distinguishing it from CIT, which is driven by sympathetic tone. Finally, BAT-specific arteriovenous metabolomics identifies glucose as the predominant circulating fuel for DIT. These findings uncover distinct molecular and metabolic features of DIT, highlighting opportunities to harness BAT activity for treating obesity and metabolic diseases without requiring cold exposure.

physiology↗

bioRxiv2025

An evolutionary leap in plant reproduction: From archegonia and motile sperm to embryo sacs and pollen tubes

The transition of plants from water to land marks a pivotal step in their evolutionary history, with the change in reproductive mode being the most fundamental. This change freed plant reproduction from aquatic dependence, fostering terrestrial lifes prosperity. Key to the shift from archegonial to embryo sac reproduction are pollen tube emergence and female gametophyte changes, though their evolutionary trajectories and mechanisms remain unclear. Ginkgo, a basal gymnosperm, occupies a pivotal position in this transition, retaining motile sperm, simplified archegonia, and primitive pollen tubes, making it an ideal model. To study how the transition between these two reproductive modes occurred, we collected transcriptome data from the reproductive organs of Ginkgo and evolutionarily representative species. Our findings reveal the MYB98-CRP-ECS module- critical for angiosperm pollen tube guidance--in Ginkgos mature archegonia, with pollen tubes containing guidance components. While egg cells are transcriptionally and functionally conserved across land plants, changes in the cell fate of other female gametophyte component cells drove the shift from archegonia to embryo sacs. Pollen tubes may originate from fern male gametophyte internalization and modification. These findings clarify the evolutionary path and molecular basis of this major reproductive transition.

plant biology↗

bioRxiv2025

OmniCellAgent: Towards AI Co-Scientists for Scientific Discovery in Precision Medicine

Real-world biomedical scientific discovery operates as an iterative lifecycle integrating four core pillars: the targeted identification and analysis of question-specific omics datasets; the context-aware interpretation of molecular data using rich biomedical prior knowledge database; comprehensive literature reviews; and the subjective creativity and expert intuition of human scientists. Together, these pillars drive robust evidence synthesis and novel hypothesis generation. The recently reported AI agents support automated omics analysis and literature review, which typically require users to predefine and curate disease-specific datasets, which is a process that remains challenging and time-consuming. In this study, we present OmniCellAgent, a novel multi-agent AI framework built on large-scale single-cell RNA sequencing (scRNA-seq) datasets, and can autonomously retrieve and analyze disease and control-related scRNAseq datasets of diverse cell types across tissues and conditions. Moreover, it incorporates a biomedical prior knowledge and literature review agents, and disease domain-specific expert agents to systematically annotate omic data-derived targets. By aggregating evidence across agents, the framework generates structured analytical reports and potential scientific hypotheses. We evaluated OmniCellAgent across multiple disease settings, demonstrating its ability to identify relevant datasets, generate omic data analysis results, and produce structured reports and scientific hypotheses. Our findings demonstrate that multi-agent AI systems lower the technical barriers to omics-driven research, thereby accelerating scientific discovery and hypothesis generation in biomedical research and precision medicine. The source code is publicly available at: https://github.com/FuhaiLiAiLab/OmniCellAgent.

bioinformatics↗