Biology subjects

Bai, L.

Publications and source records attributed to Bai, L..

At least 19 recordsLinked to original sources

bioRxiv2026

NeuroPack: Advanced Learnable Packaging Toolkit Towards Ultra-High Compression Ratio for Volumetric Neuroimaging

High-throughput 3D microscopic imaging generates brain-volume datasets at terabyte to petabyte scale, posing significant storage challenges for large-scale neuroimaging data. In this study, we introduce NeuroPack, a hyperprior-driven 3D deep-learning-based packaging toolkit specifically designed for compressing 3D microscopic volumes at the mesoscale. NeuroPack encodes a compact hyperprior stream that conditions its entropy model on local voxel-wise variance, adaptively allocating bits to preserve mesoscale anatomical details for an optimal rate-distortion balance. To ensure reproducible deployment across heterogeneous microscopy datasets and computing environments, NeuroPack implements model integerization for the hyperprior side-information pathway, whose hyper-decoder predicts the Gaussian mean and scale parameters used by the arithmetic coder. By replacing platform-dependent floating-point hyper-decoding with deterministic fixed-point integer convolutions, NeuroPack produces bit-exact entropy models across CPUs, GPUs, local workstations, and cloud servers, thereby preventing decoding drift caused by floating-point rounding, fused-multiply-add ordering, or library-specific kernels. This integerized design preserves rate-distortion behavior while maintaining efficient encoding and decoding for large anisotropic 3D microscopy volumes. Our method enables high-fidelity microscopy reconstruction at 100- to 1000-fold compression across diverse modalities and species. Quantitatively, NeuroPack delivers the strongest overall rate-distortion performance in both BD-PSNR and BD-MSSSIM, with average relative gains of 73.4% and 50.8%, respectively, over the best competing codec. Importantly, NeuroPack preserves image fidelity required for downstream morphological reconstruction and analysis, as validated by simultaneous human annotation of original and decompressed confocal whole-brain zebrafish microscopy images. Together, NeuroPack provides a learning-based compression framework explicitly tailored for microscopy data, establishing a scalable storage solution for data-intensive neuroimaging and AI-empowered biological discovery.

bioengineering↗

bioRxiv2026

Refactored intracellular domains enhance CAR expression and function

Chimeric antigen receptor (CAR) intracellular domains (ICDs) couple antigen recognition to T-cell activation. Here we worked with AI agents to refactor CAR ICDs, proposing designs and refining them as measurements accumulated over three rounds. This approach yielded KRGR-{zeta}1-40, a 45-amino-acid ICD less than a third the length of the clinical 4-1BB-CD3{zeta} (BB{zeta}) ICD. Compared with BB{zeta}, KRGR-{zeta}1-40 increased CAR surface expression and antigen-driven IFN-{gamma} release, without detectable antigen-independent IFN-{gamma} release. KRGR-{zeta}1-40 extended CAR expression and target-cell killing after transient mRNA delivery, supported the expression of four CARs from a shared mRNA dose, and enhanced B-cell depletion after lentiviral and lipid-nanoparticle delivery. These results show that iterative agent-guided design with experimental feedback can yield compact ICDs that enhance CAR expression and sustain CAR T-cell function.

bioengineering↗

bioRxiv2026

Scaling genetic discovery through automated post-GWAS interpretation

Genome-wide association studies (GWAS) have made association discovery systematic and cumulative, yet evidence for interpreting associated loci remains selective, heterogeneous and difficult to accumulate across studies. Here we present Post-GWAS Intelligence (PGI), which couples context-dependent post-GWAS analysis to a common representation of analysis-derived evidence. Its multi-agent engine, VariantAgent, executes workflows, records tool-derived results in standardized variant-centred units and generates trait-level interpretations linked to their analytical basis and evidential status. Complementary benchmarks showed that VariantAgent outperformed tested general-purpose and biomedical agents, with its largest advantage when success required both a correct answer and a valid analytical trace. Across 34 published GWAS, VariantAgent achieved a mean per-study recovery of 82.1% for directly comparable findings. The largest gains lay downstream of association mapping, where uniform analysis extended published interpretations and identified regulatory relationships at established risk loci not previously reported in the source studies. Analysis of 1,041 GWAS yielded 530,949 evidence units and corresponding reports, which PGI organized into a cross-trait catalogue. Catalogue interrogation resolved an apparent cross-trait overlap into distinct signals and trait-specific biological hypotheses. This work demonstrates how PGI can extend genetic discovery beyond association mapping by enabling biological evidence to accumulate across studies.

genetics↗

bioRxiv2026

ATR is a master regulator of the oocyte meiotic prophase checkpoint

During meiotic prophase, programmed double-strand breaks (DSBs) initiate meiotic recombination, and defects in this process result in persistent DSBs and chromosome asynapsis, triggering checkpoint-mediated oocyte elimination before primordial follicle formation. However, how the oocyte meiotic prophase checkpoint is regulated remains unclear. ATR is a DNA damage response kinase essential for male meiosis, but its role in female meiosis is poorly characterized. Here, we identify ATR as a master regulator of the oocyte meiotic prophase checkpoint. ATR loss fully rescues the survival of Dmc1-/- oocytes, which exhibit persistent DSBs and extensive chromosome asynapsis. ATR loss also fully rescues the survival of asynaptic Spo11-/- oocytes and Trip13-/- oocytes with aberrant checkpoint activation, demonstrating an essential role for ATR in the chromosome asynapsis checkpoint. Mechanistically, BRCA1 is required for ATR localization to unsynapsed chromosome axes, whereas ATR is dispensable for BRCA1 localization, establishing a BRCA1-ATR regulatory hierarchy at unsynapsed chromosome axes that is conserved between oocytes and spermatocytes. Together, these findings place ATR at the center of the oocyte meiotic prophase checkpoint and define a BRCA1-ATR signaling axis that links chromosome asynapsis to oocyte elimination.

developmental biology↗

bioRxiv2026

Local fluidization of an active cytoplasmic gel partitions large cells

Early animal embryos undergo rapid cleavages that partition cytoplasmic volumes orders of magnitude larger than those of somatic cells1. Each division must reposition nuclei and centrosomes and distribute organelles within minutes, over distances up to hundreds of micrometers2. Cleavage furrows are positioned by microtubule asters3,4, but the mechanical mechanism for long-range transport of cytoplasmic components before cytokinesis was unknown. Here, we show that cytoplasm behaves as a locally switchable active material. Fluidization at the midplane allows bulk actomyosin to convert a local mechanical asymmetry into directed global flows of all components as a composite material. Using an actin-intact cycling Xenopus egg extract together with Xenopus and medaka embryos, we find that F-actin mechanically couples microtubule asters, organelles, nuclei and centrosomes into a gel-like composite that propagates forces over hundreds of micrometers. After mitosis, Aurora B kinase patterns a locally fluidized midplane, from which myosin-II contractility drives coherent cytoplasmic flows. A fluid dynamics model accounts for the observed flow geometry and rates. Our results reveal how local control of the material state of cytoplasm converts mitotic symmetry breaking into long-range intracellular transport and identify bulk actomyosin as the active stress generator that partitions embryonic cytoplasm as a composite gel.

cell biology↗

bioRxiv2026

CENO: A Genome-Scale World Model for Evolutionary Sequence Interpretation and Programmable Regulatory Design

DNA encodes biological function across a continuum of sequence scales, from single-nucleotide and motif-level grammar to regulatory neighborhoods, chromatin-scale organization and evolutionary constraint. A useful model of genomes should therefore do more than classify short sequence windows: it should maintain nucleotide-resolution state over long contexts, score counterfactual mutations, condition on homologous sequence evidence and generate candidates that can be evaluated against structural or functional objectives. We define such a system operationally as a genomic world model: a general-purpose generative model of genome sequence space that unifies sequence understanding and sequence design through a shared state and likelihood interface. Here we introduce CENO, a family of long-context generative genomic world models designed to preserve local DNA grammar while extending usable context to regulatory and chromatin scales. CENO combines Mamba sequence-mixing layers, sparse attention layers and mixture-of-experts capacity in a single autoregressive backbone, and is trained at 300M, 600M and 1B parameter scales with a staged curriculum that progresses from 8k-token cross-domain genomic pretraining to 131k- and 1M-token whole-genome long-context continuation. We evaluate CENO under a unified world-model benchmark paradigm spanning retrieval, representation, counterfactual perturbation, reconstruction, evolutionary conditioning and design. CENO retains practical long-context inference and retrieves distal sequence in synthetic assays. In zero-shot long-context analyses, without task-specific fine-tuning, long-context continuation yields annotation- and chromatin-boundary-associated attention patterns and frozen-state representations that generalize across human cell types and mouse cell or tissue settings. To incorporate evolutionary information, we further post-train CENO on packed real multiple-sequence-alignment contexts and score variants by reference-mutant likelihood deltas, improving matched variant-effect prediction and producing evolutionary enrichment signals across species. Complementing these perturbation-based variant tests, we evaluate zero-shot long-sequence generation by partial-gene continuation, asking whether the model can recover withheld gene-scale sequence structure across eukaryotic, bacterial and archaeal species; recovery improves with model scale and later whole-genome long-context training. Finally, we use CENO as the backbone for a cell-type-specific enhancer design workflow in mouse cortex, coupling a CENO-based accessibility oracle with conditional supervised fine-tuning and oracle-guided reinforcement learning. Together, CENO provides a genome-scale sequence world-model framework for sequence interpretation, evolutionary reasoning, gene-scale reconstruction and programmable regulatory sequence generation.

genomics↗

bioRxiv2026

Functional coupling between the a4 - a5 loop and allosteric site on MKP5 is a critical determinant of catalysis

Dual-specificity phosphatases (DUSPs) that inactivate mitogen-activated protein kinases (MAPKs) are called MAPK phosphatases (MKPs). The stress-responsive MKP, MKP5, dephosphorylates p38 MAPK and c-Jun NH2-terminal kinase (JNK). Previously, we identified an allosteric compound that binds to the site within the phosphatase domain of MKP5 that is critical for the binding and dephosphorylation of both p38 MAPK and JNK. The allosteric site, comprised of the 4-5 loop, is an essential region for transmitting MAPK binding to the catalytic site. Here, we examine the contribution of additional structural rearrangements that occur upon allosteric site engagement. We show that binding of an inhibitor that engages the MKP5 allosteric site induces conformational changes in the 4-5 loop. Mutants of residues in this loop inhibited enzymatic activity, and some mutants exhibited changes in dynamics, indicating that this loop has a significant structural and dynamic role in controlling MKP5 catalysis. Enzymatic and NMR studies supported the interpretation that conformational changes and dynamics in the 4-5 loop are required for enzymatic function. Additionally, alanine mutants of R442 (4-5 loop), the catalytic base D377 ({beta}4-2 loop), and Q409A and S413A ({beta}5-3 loop) disrupted catalytic activity. These results highlight the 4-5, {beta}4-2, and {beta}5-3 loops as structurally, dynamically, and functionally interconnected for communication between the allosteric and enzymatic sites.

biochemistry↗

bioRxiv2026

MolClaw: An Autonomous Agent with Hierarchical Skills for Drug Molecule Evaluation, Screening, and Optimization

Computational drug discovery, particularly the complex workflows of drug molecule screening and optimization, requires orchestrating dozens of specialized tools in multi-step workflows, yet current AI agents struggle to maintain robust performance and consistently underperform in these high-complexity scenarios. Here we present MolClaw, an autonomous agent that leads drug molecule evaluation, screening, and optimization. It unifies over 30 specialized domain resources through a three-tier hierarchical skill architecture (70 skills in total) that facilitates agent long-term interaction at runtime: tool-level skills standardize atomic operations, workflow-level skills compose them into validated pipelines with quality check and reflection, and a discipline-level skill supplies scientific principles governing planning and verification across all scenarios in the field. Additionally, we introduce MolBench, a benchmark comprising molecular screening, optimization, and end-to-end discovery challenges spanning 8 to 50+ sequential tool calls. MolClaw achieves state-of-the-art performance across all metrics, and ablation studies confirm that gains concentrate on tasks that demand structured workflows while vanishing on those solvable with ad hoc scripting, establishing workflow orchestration competence as the primary capability bottleneck for AI-driven drug discovery.

bioinformatics↗

bioRxiv2026

Spatiotemporal organization of dynein in bulk cytoplasm promotes aster growth and positioning in large embryos

During cleavage in vertebrates, large microtubule asters grow from centrosomes in anaphase. The microtubule motor dynein pulls on these asters in bulk cytoplasm and positions centrosomes to determine cleavage geometry. However, how aster growth and pulling are coordinated remains unclear. We discovered that small metaphase asters generate a halo-like enrichment of cytoplasmic dynein at the aster periphery in medaka early embryos. In anaphase, dynein relocates from the halo to growing asters and neighboring membranous organelles, coincident with aster expansion and centrosome movement. Dynein inhibition or localized halo disruption prevents centrosome movement. Unexpectedly, dynein inhibition also suppresses aster growth and causes ectopic cytoplasmic microtubule nucleation, which leads to ectopic furrows. We propose that inactive dynein that accumulated at the metaphase aster periphery is activated in anaphase and incorporates both microtubule nucleators and organelles into growing asters to coordinately promote aster growth and pulling for efficient centrosome positioning in rapidly-dividing, large embryos.

cell biology↗

bioRxiv2026

Structure-guided discovery and engineering of miniature CRISPR-Cas12m for epigenome editing

CRISPR-based epigenome editing represents a programmable strategy to precisely modulate gene expression, holding immense promise for therapeutic applications. However, the large size of the dCas proteins substantially impedes the delivery via adeno-associated virus (AAV) vectors. Here, through iterative bioinformatics analysis, structure-guided predictions, and functional assays, we identified and characterized PmCas12m, a novel miniature subtype V-M CRISPR-Cas12m. PmCas12m exhibited flexible 5-YTN-3 PAM-dependent recognition and robust double-stranded DNA binding properties, while lacking DNA cleavage activity, thus positioning it as an ideal tool for epigenome editing. Cryogenic electron microscopy (cryo-EM) structures of PmCas12m unveiled its unique molecular mechanism of DNA binding facilitating interference. Guided by these structural insights, we employed deep mutational scanning (DMS) and protein engineering to develop xCas12m, a hypercompact variant with highly potent and specific epigenome editing capabilities in human cells. We further constructed the xCas12m-CRISPRoff platform in a single AAV vector, which achieved durable epigenetic silencing and effective inhibition of hepatitis B virus (HBV) infection in a mouse model. Collectively, these findings establish xCas12m as a versatile epigenome editing platform with transformative potential for treating diseases, paving the way for clinical translation of epigenetic therapies.

bioengineering↗

bioRxiv2026

SCALE: Scalable Conditional Atlas-Level Endpoint transport for virtual cell perturbation prediction

Virtual cell models aim to enable in silico experimentation by predicting how cells respond to genetic, chemical, or cytokine perturbations from single-cell measurements. In practice, however, large-scale perturbation prediction remains constrained by three coupled bottlenecks: inefficient training and inference pipelines, unstable modeling in high-dimensional sparse expression space, and evaluation protocols that overemphasize reconstruction-like accuracy while underestimating biological fidelity. In this work we present a specialized large-scale foundation model SCALE for virtual cell perturbation prediction that addresses the above limitations jointly. First, we build a BioNeMo-based training and inference framework that substantially improves data throughput, distributed scalability, and deployment efficiency, yielding 12.51* speedup on pretrain and 1.29* on inference over the prior SOTA pipeline under matched system settings. Second, we formulate perturbation prediction as conditional transport and implement it with a set-aware flow architecture that couples LLaMA-based cellular encoding with endpoint-oriented supervision. This design yields more stable training and stronger recovery of perturbation effects. Third, we evaluate the model on Tahoe-100M using a rigorous cell-level protocol centered on biologically meaningful metrics rather than reconstruction alone. On this benchmark, our model improves PDCorr by 12.02% and DE Overlap by 10.66% over STATE. Together, these results suggest that advancing virtual cells requires not only better generative objectives, but also the co-design of scalable infrastructure, stable transport modeling, and biologically faithful evaluation.

cell biology↗

bioRxiv2025

ZIKV NS5-mediated STAT2 degradation as a determinant of immune evasion and pathogenicity

Type I interferon (IFN) signaling is a central antiviral defense, with STAT2 driving the expression of IFN-stimulated genes (ISGs) to restrict viral infection. Many flaviviruses, including Zika virus (ZIKV), evade this pathway through non-structural protein 5 (NS5)-mediated ubiquitination and proteasomal degradation of STAT2, yet the contribution of this immune evasion strategy to viral fitness and pathogenesis remains incompletely defined. Here, using a multi-step mutational scanning strategy that jointly assessed STAT2 degradation and viral RNA replication, we identified a single NS5 residue, L162, as uniquely permissive to substitution that abolishes STAT2 degradation without compromising intrinsic replication functions. Substitution of L162 with alanine or glycine (L162A or L162G) selectively disrupted NS5 recruitment of the ZSWIM8-CUL3 E3 ubiquitin ligase, thereby preserving STAT2 stability and restoring IFN signaling. Recombinant ZIKV carrying these mutations exhibited reduced replication and enhanced ISG induction in human cells, defects fully rescued by STAT2 knockout. In A129 mice (type I IFN receptor deficient), mutant and WT viruses replicated comparably, but in human STAT2 knock-in mice, ZIKV-NS5L162G exhibited markedly reduced viral loads and disease. Notably, infection with ZIKV-NS5L162G elicited robust neutralizing antibody and T cell responses that conferred protection against WT challenge. Together, these findings establish NS5-mediated STAT2 degradation as a central determinant of ZIKV immune evasion, viral fitness and pathogenesis, and highlight disruption of NS5-STAT2 antagonism as a promising strategy for antiviral intervention and rational attenuation.

microbiology↗

bioRxiv2025

Cohesin Facilitates Nucleosome Invasion by Transcription Factors

Nucleosomes present a major barrier to transcription factor (TF) binding. However, a subgroup of TFs known as pioneer factors (PFs) can recognize motifs covered by nucleosomes and initiate chromatin opening. PFs also bind nucleosomal substrates with high affinity in vitro, which may facilitate their nucleosome invasion in vivo. Here, we show that LexA, a bacterial TF with poor nucleosome binding in vitro, can rapidly invade into a well-positioned nucleosome from a motif positioned at the dyad when expressed ectopically in yeast. This striking contrast between LexA binding in vitro and in vivo raises the possibility that TFs can exploit nucleosome dynamics in vivo to access occluded sites. Surprisingly, we find that LexA-mediated chromatin opening can occur in the absence of DNA replication, chromatin remodeling, histone turnover, and a few histone chaperones. Instead, nucleosome invasion by LexA and a native PF, Cbf1, is promoted by the cohesin complex, illustrating intriguing a connection between cohesin and TF binding. Together, our results demonstrate that even non-pioneer TFs like LexA can bind and displace nucleosomes in vivo, through a process facilitated by cohesin.

molecular biology↗

bioRxiv2025

scMOBA: A conversational single-cell Multi-Omics Brain Agent across species

Single-cell and spatial multi-omics are revolutionizing our understanding of the complexity in the developmental, aging, and diseased brain, but integrating this knowledge across modalities and species remains challenging. To bridge this gap, we propose scMOBA, a conversational single-cell Multi-Omics Brain Agent established by a large language model, a gene encoder, and a cross-attention projector. The scMOBA was pre-trained on 130 million single-cell and spatial multi-omics data spanning the entire brain across diverse species, development, aging and diseases. The pre-training utilized a novel multi-omics Feature-Question-Answer (FQA) paradigm, enabling the model to generate biological answers from feature inputs and textual queries. This unique scheme facilitates superior zero-shot inference capabilities without requiring additional fine-tuning. We demonstrate that scMOBA achieves state-of-the-art performance in fine-grained cell type classification across different species and modalities, as well as in batch correction and multi-omics data integration. Furthermore, scMOBA significantly boosts the accuracy of kinds of critical downstream tasks, including cell-type specific aging clock construction and disease status prediction. Overall, scMOBA serves as a powerful scientific discovery engine for multi-omics brain research, advancing the precision prediction and early intervention of neurological aging and diseases.

bioinformatics↗

bioRxiv2025

Scalable integration and prediction of unpaired single-cell and spatial multi-omics via regularized disentanglement

Deciphering cellular states requires methods capable of integrating large-scale heterogeneous single-cell and spatial omics data. However, these data are typically unpaired due to destructive assays and further confounded by modality heterogeneity, technical noise, and immense scale. Here we present scMRDR, a scalable computational framework based on regularized disentangled representation learning for integrating fully unpaired single-cell and spatial multi-omics datasets. Built on a unified and structure-preserving architecture, scMRDR removes the need for pairing supervision while maintaining computational efficiency, enabling scaling to large datasets spanning multiple disparate omics modalities. Across diverse real-world benchmarks, scMRDR demonstrates strong performance in batch correction, modality alignment, and biological signal preservation. The framework further supports cross-modal translation across omics modalities and enables spatial coordinate imputation for non-spatial single-cell datasets using a reference atlas. The resulting spatial mapping allows spatially resolved analyses, including identification of spatially variable genes and characterization of epigenetic regulatory programs in their native tissue context. These capabilities position scMRDR as a scalable and versatile framework for large-scale multi-omics integration.

genomics↗

bioRxiv2025

A nutrient-sensitive enterokine coordinates developmental plasticity through inter-organ signalling

Animal survival in fluctuating environments depends on the ability to modulate their developmental pace in response to nutrient availability, a phenomenon known as developmental plasticity. In Drosophila larvae, we uncover a critical endocrine mechanism that coordinates this process under conditions of amino acid restriction. We identify the peptide hormone Limostatin as an enterokine, produced by a small population of larval midgut enteroendocrine cells, that acts systemically to inhibit the expression and release of dIlp2, a major insulin-like peptide controlling developmental progression. Limostatin expression and secretion by enteroendocrine cells is triggered by reduced amino acid availability through an inter-organ relay involving the fat body and neuroendocrine insulin-producing cells in the brain. In turn, Limostatin participates in a feedback control loop that slows down developmental progression once systemic nutrient shortage is sensed. This bidirectional gut-brain axis enables larvae to preserve viability under nutritional stress. Our findings define the larval gut as a nutrient-sensitive endocrine organ and position Limostatin as a key regulator of developmental plasticity. Our work expands the concept of decretins to include developmental pace control, suggesting that enterokines that regulate IGF signalling, rather than insulin release per se, may represent an evolutionarily conserved or convergent strategy in regulating developmental plasticity.

developmental biology↗

bioRxiv2025

Chemo-mechanotherapy of fibrosis: dynamic control of biological materials through tissue-architecture-dependent crosslink disruption

Extracellular matrix (ECM) stiffness drives cell fate, but tissues are nonlinear and difficult to modify chemically without harming patients. Here, we show that nonlinear mechanical responses of fibrous and porous tissues can be controlled by mechanically modulating crosslinks. Through integrated in vivo, in vitro, and computational approaches, we found that physiological dynamic stretching disrupts pathological advanced glycation end-product (AGE) crosslink-induced fibrogenesis in an architecture-dependent manner. This was effective in lung-like porous scaffolds but not in liver-like fibrous matrices, revealing a critical structure-property relationship. Computational and in vitro models established that this modulates cell-ECM feedback, and that controlling mechanical crosslinking modulates key nonlinearity of networked solids. This discovery inspired a non-invasive, "mechanotherapuetic" ventilation protocol, which reversed pulmonary fibrosis in mice by physically disrupting ECM pathological crosslinks. The therapeutic efficacy was further amplified when combined with pharmacological AGE inhibition. These findings establish design principles for dynamically reprogrammable biological materials and mechano-targeted therapies.

bioengineering↗

bioRxiv2025

OriGene: A Self-Evolving Virtual Disease Biologist Automating Therapeutic Target Discovery

Here, we present OriGene, a self-evolving multi-agent system that functions as a virtual disease biologist, systematically identifying original and mechanistically grounded therapeutic targets at scale. OriGenes architecture integrates over 600 specialized tools through a Model Context Protocol (MCP), enabling it to reason across diverse data modalities including genomics, protein networks, pharmacology, clinical records and literature evidence, to generate and prioritize target discovery hypotheses. We implemented a strategy combining a knowledge graph-based Tool RAG with an advanced agent selection mechanism to enable dynamic, context-aware tool deployment. Through a self-evolving framework, OriGene continuously integrates human and experimental feedback to iteratively refine its core thinking templates, tool composition, and analytical protocols, thereby enhancing both accuracy and adaptability over time. To comprehensively evaluate its performance, we established TRQA, an original benchmark comprising over 1,900 expert-level question-answer pairs spanning a wide range of diseases and target classes. OriGene consistently outperforms human experts, leading research agents, and state-of-the-art large language models in accuracy, recall, and robustness, particularly under conditions of data sparsity or noise. Critically, OriGene nominated previously underexplored therapeutic targets for liver (GPR160) and colorectal cancer (ARG2), which demonstrated significant anti-tumor activity in patient-derived organoid and tumor fragment models mirroring human clinical exposures. These findings demonstrate OriGenes potential as a scalable and adaptive platform for AI-driven discovery of mechanistically grounded therapeutic targets, offering a new paradigm to accelerate drug development.

bioinformatics↗