Biology subjects

Jun, H.

Publications and source records attributed to Jun, H..

8 recordsLinked to original sources

bioRxiv2026

Evaluating agentic AI for biological discovery in autonomous and copilot settings

Advances in large language models (LLMs)-based artificial intelligence (AI) agents have improved their ability to execute structured analytical workflows, including standard bioinformatic pipelines for biological discovery. However, computational biology rarely consists of deterministic pipeline execution alone. Biological datasets are heterogeneous and noisy, and meaningful discovery often requires open-ended hypothesis generation and iterative reasoning over multimodal evidence. These challenges are particularly evident in multi-omic studies, where paired molecular modalities and heterogeneous clinical contexts create both opportunities and obstacles for discovery. The extent to which emerging agentic AI systems can support or automate this mode of scientific discovery remains poorly understood. Here, we systematically evaluated the capabilities and limitations of agentic AI for biological discovery using multi-omic single cell datasets spanning 11 cancer types. We developed the Multistep Multimodal Multiomic Agentic (M3A) Framework to support LLM-driven reasoning over persistent multimodal data states and to capture agentic reasoning behavior in autonomous and human-AI copilot settings. Using this framework, we assessed AI agents across complementary tasks, including autonomous cell-type annotation, generation of falsifiable biological hypotheses from gene programs, and copilot experiments testing the effect of human involvement and domain expertise. We found that current AI agents are effective at broad, systemic exploration of complex data, whereas domain experts remain critical for methodological guidance and biological synthesis across analyses. Together, our results delineate the current potential and boundaries of agentic AI in computational biology, and establish a framework for evaluating AI systems designed to support biological discovery.

cancer biology↗

bioRxiv2025

Cancer-associated metabolic alterations facilitated by astrocyte glycogenolysis mediate neuropathic pain chronicity

Chronic pain remains a major unmet medical challenge, yet the metabolic checkpoints that govern its persistence are poorly defined. Astrocytes are increasingly recognized as chemically programmable hubs that tune neuronal excitability through metabolic circuits. Building on reports that astrocyte-neuron lactate shuttling (ANLS) in the anterior cingulate cortex (ACC) supports chronic pain, we asked how astrocytic metabolic states evolve over the course of pain chronification. Using untargeted metabolomics of the ACC combined with GFAP-RiboTag-based astrocyte-specific transcriptomics, we provide a time-resolved map of astrocytic metabolism across the transition from acute nociception to chronic neuropathic pain. This analysis reveals a biphasic glycogen program--an acute glycogenolysis-triggered glycogen supercompensation--that culminates in the emergence of a Warburg-like metabolic signature with tissue acidification associated with sustained lactate shuttling and persistent circuit activation. Using glycogen phosphorylase inhibitors (GPI-1, GPI-2) as pharmacological probes, we show that early glycogenolysis blockade attenuates this Warburg-like shift, partially normalizes ACC metabolic signatures, and reduces long-lasting mechanical hypersensitivity, without impairing acute nociceptive sensitization. These findings identify astrocytic metabolic reprogramming as a pharmacologically tractable circuit-level process and nominate glycogenolysis as an upstream biochemical gate and potential therapeutic control point in neuropathic pain. HIGHLIGHTSMost experimental models of chronic pain emphasize progressive sensitization, yet few have identified a chemically tractable switch that converts transient nociception into a persistent cortical state. Here, we propose astrocytic glycogenolysis as such a switch--a gatekeeper process whose brief engagement can reconfigure neuroglial metabolism toward a Warburg-like signature that stabilizes hyperexcitable circuit dynamics. By leveraging small-molecule glycogen phosphorylase inhibitors as mechanistic precision probes, we move beyond correlation to delineate a pathway-level control architecture that can be transiently perturbed to dissociate acute nociceptive processing from pain chronification. This framework reframes astrocytes from metabolic support cells to state-setting regulators of cortical excitability and positions glycogenolysis as a therapeutically actionable leverage point. More broadly, our findings underscore a methodological opportunity: applying chemical biology to neuroglial metabolism can expose governing checkpoints that are often obscured by descriptive physiology, enabling mechanism-first, pathway-directed intervention strategies across neurological disease contexts.

neuroscience↗

bioRxiv2024

Early disruption of entorhinal dopamine in a knock-in model of Alzheimer's disease

The entorhinal cortex (EC) is a critical brain area for memory formation, while also the region exhibiting the earliest histological and functional alterations in Alzheimers disease (AD). The EC thus has been long hypothesized as one of the originating brain areas of AD pathophysiology, although circuit mechanisms causing its selective vulnerability remain poorly understood. We found that dopamine neurons projecting their axons to the lateral EC (LEC), critical for memory formation in healthy brains, become dysfunctional and cause memory impairments in early AD brains. In amyloid precursor protein knock-in mice with associative memory impairment, LEC dopamine activity and associative memory encoding of LEC layer 2/3 neurons were disrupted in parallel from the early pathological stage. Optogenetic reactivation of LEC dopamine fibers, as well as L- DOPA treatment, rescued associative learning behavior. These results suggest that dysfunction of LEC-projecting dopamine neurons underlies memory impairment in AD from early stages, pointing to a need for clinical investigation of LEC dopamine in AD patients.

neuroscience↗

bioRxiv2022

Expression profiles of hypoxia-related genes of cancers originating from anatomically similar locations using TCGA database analysis

Hypoxia is a well-recognized characteristic of the tumor microenvironment of solid cancers. This study aimed to analyze hypoxia-related genes shared by groups based on tumor location. Nine hypoxia-related pathways from the Kyoto Encyclopedia of Genes and Genomes database or the Reactome database were selected, and 850 hypoxia-related genes were analyzed. Based on their anatomical locations, 14 tumor types were categorized into the following six groups. The group-specific genetic risk score was classified as high or low risk based on mRNA expression, and survival outcomes were evaluated. The risk scores in the Female reproductive group and Lung group were internally and externally validated. In the Female reproductive group, CDKN2A, FN1 and ITGA5, were identified as hub genes associated with poor prognosis, while IL2RB and LEF1 were associated with favorable prognosis. In the Lung group, ITGB1 and LDHA were associated with poor prognosis, and GLS2 was associated with favorable prognosis. Functional enrichment analysis showed that the Female reproductive group was enriched in terms related to cilia and skin, while the Lung group was enriched in terms related to cytokines and defense. This analysis may lead to better understanding of the mechanisms of cancer progression and facilitate establishing new biomarkers for prognosis prediction.

bioinformatics↗

bioRxiv2020

Rapid Prototyping of Wireframe Scaffolded DNA Origami using ATHENA

Wireframe DNA origami assemblies can now be programmed automatically from the "top-down" using simple wireframe target geometries, or meshes, in 2D and 3D geometries using either rigid, six-helix bundle (6HB) or more compliant, two-helix bundle (2HB or DX) edges. While these assemblies have numerous applications in nanoscale materials fabrication due to their nanoscale spatial addressability and high degree of customization, no easy-to-use graphical user interface software yet exists to deploy these algorithmic approaches within a single, stand-alone interface. Here, we present ATHENA, an open-source software package with a graphical user interface that automatically renders single-stranded DNA scaffold routing and staple strand sequences for any target wireframe DNA origami in 2D or 3D using 2HB or 6HB edges. ATHENA enables external editing of sequences using the popular tool caDNAno, demonstrated here using asymmetric nanoscale positioning of gold nanoparticles, as well as atomic-level models for molecular dynamics, coarse-grained dynamics, or other computational chemistry simulation approaches. We anticipate ATHENA will significantly reduce the barrier for non-specialists to perform wireframe DNA origami sequence design and fabrication for custom applications in materials science, nanotechnology, therapeutics, and other areas.

biophysics↗

bioRxiv2019

Disrupted remapping of place cells and grid cells in knock-in model of Alzheimer’s disease

Patients with Alzheimers disease (AD) frequently suffer from spatial memory impairment and wandering behavior, but brain circuits causing such symptoms remain largely unclear. In healthy brains, spatially-tuned hippocampal place cells and entorhinal grid cells represent distinct spike patterns in different environments, a circuit function called "remapping" that underlies pattern separation of spatial memory. We investigated whether knock-in expression of mutated amyloid precursor protein deteriorates the remapping of place cells and grid cells. We found that the remapping of CA1 place cells was disrupted although their spatial tuning was only mildly diminished. Grid cells in the medial entorhinal cortex (MEC) were impaired, sending severely disrupted remapping signals to the hippocampus. Furthermore, fast gamma oscillations were disrupted in both CA1 and MEC, resulting in impaired fast gamma coupling in the MEC[->]CA1 circuit. These results point to the link between grid cell impairment and remapping disruption as a circuit mechanism causing spatial memory impairment in AD.

pathology↗

bioRxiv2019

Early Stem Cell Aging in the Mature Brain

Stem cell dysfunction drives many age-related disorders. Identifying mechanisms that initially compromise stem cell function represent early targets to enhance stem cell behavior later in life. Here, we pinpoint multiple factors that disrupt neural stem cells (NSC) in the adult hippocampus. We find that NSCs exhibit asynchronous maintenance by identifying short-term (ST-NSC) and intermediate-term NSCs (IT-NSCs). ST-NSC divide rapidly to generate neurons and deplete in the young brain. Meanwhile, multipotent IT-NSCs persist for months, but are pushed out of homeostasis by lengthening quiescence. Single cell transcriptome analysis of deep NSC quiescence revealed several hallmarks of biological aging in the mature brain and identified tyrosine-protein kinase Abl1 as an NSC pro-aging factor. Treatment with the Abl-inhibitor Imatinib increased NSC proliferation without impairing NSC maintenance in the middle-aged brain. Further intersectional analysis of mature NSC with old epidermal, hematopoietic and muscle stem cell transcriptomes identified consensus changes in stem cell aging. Our study elucidates multiple origins of adult neurogenesis decline and reveals that hippocampal NSCs are particularly vulnerable to a shared stem cell aging signature.

neuroscience↗

bioRxiv2018

Chromatic Bacteria - A broad host-range plasmid and chromosomal insertion toolbox for fluorescent protein expression in bacteria

Differential fluorescent labelling of bacteria has become instrumental for many aspects of microbiological research, such as the study of biofilm formation, bacterial individuality, evolution, and bacterial behaviour in complex environments. We designed a variety of plasmids, each bearing one of eight unique, constitutively expressed fluorescent protein genes in conjunction with one of four different antibiotic resistance combinations. The fluorophores mTagBFP2, mTurquoise2, sGFP2, mClover3, sYFP2, mOrange2, mScarlet-I, and mCardinal, encoding for blue, cyan, green, green-yellow, yellow, orange, red, and far-red fluorescent proteins, respectively, were combined with selectable markers conferring tetracycline, gentamicin, kanamycin, and/or chloramphenicol resistance. These constructs were cloned into three different plasmid backbones: a broad host-range plasmid, a Tn5 transposon delivery plasmid, and a Tn7 transposon delivery plasmid. The utility of the plasmids and transposons was tested in bacteria from the phyla Actinobacteria, Proteobacteria, and Bacteroidetes. We were able to tag representatives from the phylum Proteobacteria at least via our Tn5 transposon delivery system. The here constructed plasmids are available to the community and provide a valuable tool to investigate bacteria-bacteria, bacteria-host, and bacteria-environmental interactions.

microbiology↗