Biology subjects

Zhang, B.

Publications and source records attributed to Zhang, B..

At least 109 records · Page 6Linked to original sources

bioRxiv2025

GlycoTraitR: an R package for characterizing structural heterogeneity in N-linked glycoproteomics data

Glycoproteomics data are rapidly accumulating due to advances in mass spectrometry instrumentation and the development of specialized search engines (e.g., pGlyco3, Glyco-Decipher) that enable identification of N-linked glycopeptide spectral matches (GPSMs) together with glycan structures. These advances have greatly expanded the scale and depth of N-linked glycopeptides; however, the intrinsic structural heterogeneity of glycosylation remains challenging to interpret. No existing tool provides a unified trait-based framework for analyzing N-linked GPSM data at both the glycosylation-site and protein levels. We developed glycoTraitR, an R package for trait-based analysis of structural heterogeneity in N-linked glycoproteomics data. GlycoTraitR provides a unified workflow to import GPSMs from search engine outputs, extract biologically interpretable glycan structural traits, and perform comparative analyses of micro- and macro-heterogeneity across experimental conditions using statistical testing. ImplementationThe R package and the source code of glycoTraitR are freely available on github at https://github.com/matsui-lab/glycoTraitR. A more detailed introduction and quick start guide are avaible at https://matsui-lab.github.io/glycoTraitR/.

bioinformatics↗

bioRxiv2025

IGHV3-53 antibody abundance drives divergent SARS-CoV-2 immune imprinting

The mechanisms driving divergent SARS-CoV-2 immune imprinting in populations with distinct SARS-CoV-2 exposure histories remain unclear. However, conventional wild-type mouse models fail to recapitulate severe imprinting, even after potent ancestral-strain mRNA immunization, hindering imprinting-related mechanistic investigation and vaccine evaluation. Here, we surprisingly found that V(D)J-humanized mice could faithfully recapitulate human severe imprinting phenotypes. Comprehensive antibody repertoire analysis and deep mutational scanning-based epitope mapping of 583 monoclonal antibodies from these models revealed that the pre-existing IGHV3-53/66 antibody abundance determine imprinting severity through antibody-mediated masking of Omicron-specific epitopes. Both passive IGHV3-53/66 antibody transfer and IGHV3-53 knock-in were sufficient to induce severe imprinting in wild-type mice. Together, these findings demonstrate that the V(D)J germline repertoire--even a single germline-encoded antibody response--can profoundly shape humoral imprinting severity. Accordingly, we established an IGHV3-53 knock-in mouse model that accurately recapitulates the human antibody landscape, providing a valuable tool for guiding future COVID-19 vaccine updates.

immunology↗

bioRxiv2025

GlycanGT: A Foundation Model for Glycan Graphs with Pretrained Representation and Generative Learning

MotivationGlycans are highly diverse biological sequences, but their functional understanding has lagged behind that of proteins and nucleic acids. Many glycans remain incompletely characterized or ambiguously annotated, limiting computational analyses. Existing computational approaches are primarily graph-based, capturing local structural features but struggling to model global patterns and incomplete sequences. ResultsWe present GlycanGT, a foundation model for glycans built on a graph transformer architecture. Glycans were represented as graphs with monosaccharides as nodes and glycosidic bonds as edges, and the model was pretrained using a masked language modeling objective. GlycanGT demonstrated higher performance than existing methods across 8 benchmark classification tasks (e.g., 0.734 Macro-F1 in domain prediction and 0.844 AUPRC for immunogenicity classification), and its embeddings formed biologically meaningful clusters that recovered known N- and O-glycan categories. Moreover, GlycanGT accurately proposed candidates for ambiguous sequences, maintaining >80% top-5 accuracy for both monosaccharide and glycosidic bond predictions under high masking levels. Availability and implementationThe pretrained GlycanGT model weights and usage scripts are available on Hugging Face: https://huggingface.co/Akikitani295/GlycanGT. Additional scripts used for analyses in the paper are publicly available on GitHub: https://github.com/matsui-lab/GlycanGT. Contact: matsui.yusuke.d4@f.mail.nagoya-u.ac.jp

bioinformatics↗

bioRxiv2025

Cilia.io: Computer vision and machine learning reveal spatial patterns of cilia beating dynamics in the spinal cord

Motile cilia coordinate fluid flows that are essential for normal tissue physiology and function. Cilia display diverse beating waveforms, and while pronounced defects are strongly associated with motile ciliopathies, more subtle alterations may also influence disease manifestations. Finer quantification of ciliary dynamics is critical for a full understanding of cilia-associated disorders, but the heterogeneity of cilia beating dynamics makes accurate and robust characterization challenging. Because existing tools have proven to be limiting in noisy in vivo environments, we developed Cilia.io, a machine learning (ML)-based quantification tool that uses state-of-the-art vision transformers to segment cilia out of the background based on their biological features. Cilia.io enables fast, accurate, and reproducible quantification of motile cilia morphodynamics and outperforms existing tools. Indeed, using Cilia.io, we discovered distinct regional differences in ciliary waveforms in the zebrafish spinal cord. The ability of Cilia.io to capture subtle ciliary defects was further demonstrated by analyzing a novel allele in the ciliopathy gene bbs2 that causes a highly heterogeneous scoliosis phenotype. In these mutants, only dorsal cilia displayed altered beating dynamics, while ventral cilia remained largely unaffected. Our new tool therefore represents a substantial advance on existing methods and suggests that additional fine-scale analyses of ciliary beating will be important for understanding organismal phenotypes and cilia-driven disease.

developmental biology↗

bioRxiv2025

Integration of single-cell transcriptome and genetic profiles reveals critical cell types and genes in inflammatory bowel disease

The progression of inflammatory bowel disease (IBD), grouped into Crohns disease (CD) and ulcerative colitis (UC), involves multiple cell types and genes, but the identification of critical cell types and genes that contribute to intestinal inflammation remains challenging. By integrating human intestinal single-cell transcriptomic data from approximately 200 donors, we revealed the stromal cell-derived network in CD, enrichments of immune cells such as Th1 in CD, atypical B cells and Tfh in UC, AXL+SIGLEC6+ DCs and IgG+ Plasma B cells in both, and the association between SPP1+ Macrophages, FAP+ Fibroblasts and anti-TNF treatment resistance. We also investigated the gene expression characterization of enteric neurons. Additionally, by linking cell-type-specific transcriptional alterations with genetic profiles of disease risk and intestinal gene expression, we proposed B cells and cycling epithelial cells as critical cell types, and identified epithelial genes, such as GPX4, as associated with CD pathology. In summary, our study comprehensively characterized the single-cell landscape of human intestines in IBD and provided insights into cell types and genes highly involved in genetically determined CD pathology, thereby informing CD prediction, diagnosis, and the development of new therapeutic targets.

immunology↗

bioRxiv2025

Efficient Identification of Short Tandem Repeats via Context-Aware Motif Discovery and Ultra-Fast Sequence Alignment

Tandem repeats (TRs) are highly polymorphic genomic elements, associated with diverse molecular traits and implicated in numerous human diseases. However, large-scale analysis of TRs has been limited by computational challenges, including motif recognition, detection in complex regions, and excessive computational cost. Here we present FastSTR, a computationally efficient tool for precise detection and characterization of TRs. FastSTR integrates a context-aware N-gram motif model with a segmented global alignment algorithm to enable accurate motif identification and boundary definition, even for repeat units up to 8 bp. Across 13 species, FastSTR achieved >90% recall and 99% precision, running several times faster than existing methods white outperforming them in both sensitivity and accuracy. Applied to the human genome, FastSTR uncovered previously unannotated HSATII elements, resolved population-specific TR demonstrate, and identified recurrent STR alterations in lung cancer. These results demonstrate FastSTR as a versatile framework for TR annotation and discovery, advancing studies of genome evolution, genetic diversity, and disease.

bioinformatics↗

bioRxiv2025

A unified language model bridging de novo and fragment-based 3D molecule design delivers potent CBL-B inhibitors for cancer treatment

The rational design of small molecules is central to drug discovery, yet current artificial intelligence (AI) methodologies for generating three-dimensional (3D) molecules are often siloed, focusing on either de novo design or fragment-based design. The lack of a holistic framework limits AIs application across the complex and multi-step pipeline spanning from novel scaffold identification to lead compound optimization, and prevents AI from effectively learning from the entire process. Here, we introduce UniLingo3DMol, a language model for 3D molecular generation, empowered by fragment permutation-capable molecular representation alongside multi-stage and multi-task training strategy. This integrated design enables UniLingo3DMol to seamlessly span both de novo and fragment-retained molecular design, demonstrating superior performance over existing generation models in in silico evaluations across more than 100 diverse biological targets. We further leveraged UniLingo3DMol in the design of inhibitors targeting CBL-B, a crucial immune E3 ubiquitin ligase and attractive immunotherapy target. This strategy led to a lead compound demonstrating excellent in vitro activity and robust in vivo anti-tumor efficacy. Our findings establish UniLingo3DMol as a generalized and powerful platform, showing the strong potential to advance AI-driven drug discovery.

bioinformatics↗

bioRxiv2025

Bioengineered 3D hPSC-cholangiocyte ducts with physiological signals for biliary disease modelling

The progression of intrahepatic biliary diseases remains poorly understood, underscoring the urgent need to develop physiologically relevant human intrahepatic cholangiocyte disease models. Current approaches lack the complexity and throughput to capture how diverse biliary microenvironmental signals shape cholangiocyte behavior. To address this gap, we created a fully epithelialized and perfusible 3D bile duct from human pluripotent stem cell-derived cholangiocytes characterized by robust primary ciliation, apical-basal polarity and CFTR-mediated chloride conductance. Our results revealed physiologically relevant fluid flow and biliary stroma cells as essential components for sustaining cholangiocyte epithelial barrier integrity and ciliation. From here, we interrogated a broad spectrum of biliary signals to study bile acid toxicity and cytokine-driven injury, offering an unprecedented view into intrahepatic cholangiocyte stress responses and potential pathological mechanisms. By integrating physiologically relevant signals, fidelity and functional resolution, this platform provides the foundation needed to precisely decode pathogenic drivers and accelerate therapeutic development for devastating cholangiopathies. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=198 HEIGHT=200 SRC="FIGDIR/small/686855v1_ufig1.gif" ALT="Figure 1"> View larger version (59K): org.highwire.dtl.DTLVardef@11c1255org.highwire.dtl.DTLVardef@1f78d11org.highwire.dtl.DTLVardef@14e9398org.highwire.dtl.DTLVardef@1bc08d5_HPS_FORMAT_FIGEXP M_FIG C_FIG

cell biology↗

bioRxiv2025

NEAT-DNA: A Chemically Accurate, Sequence-Dependent Coarse-Grained Model for Large-Scale DNA Simulations

DNAs physical properties play a central role in genome organization and regulation, but simulating its behavior across biologically relevant scales remains a major computational challenge. Coarse grained DNA models have enabled faster simulations, yet they often sacrifice chemical accuracy or produce unphysical conformations, limiting their utility for studying genome structure. A key difficulty has been constructing a model that is both efficient enough for large-scale simulations and faithful to the molecular mechanics of DNA. Here we introduce NEAT-DNA, a new coarse-grained DNA model that resolves longstanding limitations in physical realism and parameter optimization. By combining a physically principled energy formulation with a unified training framework that integrates data from both atomistic simulations and experiments, NEAT-DNA accurately reproduces sequence-dependent structure and flexibility while remaining computationally efficient. This approach marks a significant advance over previous models, which either lacked sequence specificity or introduced distortions inconsistent with experimental observations. NEAT-DNA bridges this gap, offering a high-fidelity yet tractable representation of DNA suitable for exploring chromatin folding. More broadly, it provides a foundation for large-scale simulations that couple molecular detail with gene-level chromatin organization, opening new avenues for predictive modeling in structural genomics.

biophysics↗

bioRxiv2025

Circulating cell-free RNA reflects inflammatory and airway remodeling signatures in an equine model of asthma

Circulating cell-free RNA (cfRNA) offers a minimally invasive means to monitor dynamic gene expression across tissues, providing a promising source of biomarkers for inflammatory diseases. Equine asthma, a prevalent respiratory disorder, represents a valuable translational model for the human condition. Here, we present the first transcriptomic analysis of plasma cfRNA from asthmatic and healthy horses, alongside comparative profiling of bronchoalveolar lavage and whole blood. Tissue-of-origin deconvolution revealed asthma-associated immune cell redistribution, characterized by reduced B-cell and increased T-cell signatures that were undetectable in the conventional sampling. Differential expression analysis identified up-regulation of key alarmins and transcripts linked to airway remodeling and immune activation. Cross-compartment correlation demonstrated that cfRNA captures distinct yet biologically relevant molecular signatures consistent with asthma-related pathways. These results establish cfRNA as a sensitive complementary approach for characterizing molecular perturbations in asthma pathogenesis and highlight its potential for biomarker discovery across inflammatory and allergic airway diseases.

genomics↗

bioRxiv2025

YAP1 defines an emergent, plastic population of relapsed small cell lung cancer

Small cell lung cancer (SCLC) is an aggressive neuroendocrine malignancy characterized by rapid onset of chemoresistance and poor clinical outcomes. Transcriptional heterogeneity among treatment-naive SCLC tumors underlies four transcriptional subtypes, each with distinct clinical vulnerabilities. Though previously hypothesized to delineate a distinct subtype, expression of YAP1 is largely absent from treatment-naive, pure SCLC. To characterize relapsed SCLC, circulating tumor DNA, circulating tumor cells, and core needle biopsies from SCLC patients and preclinical models following resistance to standard-of-care therapies were analyzed. In contrast to treatment-naive SCLC, these analyses reveal an emergent YAP1-positive cell population that coincides with treatment resistance. These YAP1-positive cells exhibit characteristics of drug tolerant persister cells, including senescence, stemness, and plasticity, as YAP1 positive cells largely abandon features characteristic of SCLC to adopt those of large-cell neuroendocrine carcinoma (LCNEC). As a result of this SCLC-like to LCNEC-like evolution, YAP1-positive cells lack several clinically relevant SCLC surface targets (i.e., DLL3, SEZ6), but are enriched for others (i.e., B7-H3, TROP2). We propose a model where YAP1 expressing cells emerge with SCLC treatment resistance and characterize a tenacious subpopulation capable of diverging from the treatment naive lineage and adopting features to evade therapeutic response.

cancer biology↗

bioRxiv2025

High-Resolution Characterization of rAAV Genomes for Vector Quality Using Long-Read Sequencing

Recombinant adeno-associated virus (rAAV)-mediated gene therapy has been applied for human diseases. However, the rAAV capsids contain heterogeneous mixtures of full-length and truncated genomes and, depending on the manufacturing process, residual host cell and plasmid DNA. Therefore, a method is needed to characterize the encapsidated DNA of rAAV in order to support process development and batch release. The emerging long-read sequencing (LRS) has achieved AAV single-genome resolution. Here we propose a Python-based LRS profiling framework to classify and quantitate residual DNA species in rAAV products. We designed a reference that contains universal genetic components that are commonly used in rAAV production, including AmpR, KanR, Rep and Cap genes along with HPV18, Ad5 and hg38 genomes. We accessed the impurities of rAAV production from public and in-house LRS datasets. Analyzing the lambda fragments supplemented in these datasets showed that sequencing introduced size biases, which couldnt be fully corrected by regression but is improvable within library preparation. Functional potential of impurities were assessed through indicators derived from long-read alignments, which enabled us to quantitatively compare impurities between manufacturing batches. We demonstrated that LRS provides informative metrics for rAAV production and can facilitate process development to ensure therapeutic product safety and quality.

genomics↗

bioRxiv2025

Inducing TRIB2 Targeted Protein Degradation to Reverse Chemoresistance in Acute Myeloid Leukemia

The myeloid oncogene TRIB2 is a key driver of acute myeloid leukaemia (AML) pathogenesis, promoting chemoresistance and blocking differentiation through ubiquitin-mediated degradation of the C/EBP transcription factor. Despite its stable and sometimes elevated expression across AML subtypes, TRIB2 remains a clinically-untargeted vulnerability. Here, we present a comprehensive investigation into TRIB2 degradation mechanisms using multimodal approaches, including CRISPR knockout, mutational protein stability, small molecule TRIB2 engagement and evaluation of a novel targeted protein degrader (TRIB2-PROTAC). We identify Afatinib, a multi-ERBB covalent inhibitor, as a rapid inducer of TRIB2 degradation, triggering AML cell death via an ERBB-independent pathway. Importantly, TRIB2 degradation synergized with cytarabine, the frontline AML chemotherapy, amplifying therapeutic efficacy. Mapping of TRIB2 ubiquitination sites revealed Lys-63 as critical for its own proteolytic turnover, and a Lys to Arg degradation-resistant mutant (KallR) conferred enhanced chemoresistance and increased leukaemic engraftment in vivo. CRISPR-mediated TRIB2 knockout validated an essential role in AML cell survival. Consistently, the novel TRIB2-PROTAC (compound 5K) achieved robust TRIB2 degradation and AML cell killing at low micromolar concentrations. These findings establish TRIB2 as a compelling therapeutic target in AML and demonstrate that leveraging the ubiquitin-proteasome system to degrade TRIB2 offers a promising strategy to overcome chemoresistance. This work provides strong preclinical rationale for the development of TRIB2-targeting therapies in AML.

cancer biology↗

bioRxiv2025

GABA-GABA_A Receptor Signaling Orchestrates Invasion and Metastasis in Triple Negative Breast Cancer

Cancer is a leading cause of death globally, with the majority of cancer-related deaths resulting from cancer metastasis -the process by which cancer cells disseminate to distant sites. To metastasize, cancer cells acquire traits in support of diverse cellular processes that enable dissemination, survival, and colonization. Tumor cell dissemination requires invasion at local and distant sites and this process can be influenced by intrinsic and extrinsic factors. Here, we investigate the role of the neurotransmitter gamma-aminobutyric acid (GABA) in triple-negative breast cancer (TNBC) invasion and metastasis. TNBC cells increased invasion in response to GABA and this was found to be mediated through the GABAA receptor family. TNBC cell lines were found to be responsive to exogenous GABA and also produced endogenous GABA. Pharmacological inhibition of GABAA receptors reduced TNBC invasion and cancer cell dissemination and resulted in inhibition of GSK3 activity. TNBC cell lines were found to express the GABRE subunit and loss of GABRE impaired GABA-mediated invasion and tumor cell dissemination. These findings support a role for GABA signaling through GABAA receptors in mediating TNBC progression.

cancer biology↗

bioRxiv2025

Vaccine Elicitation of HIV-1 Neutralizing Antibodies Against Both V2 Apex and Fusion Peptide in Rhesus Macaques

Broadly neutralizing antibodies targeting multiple sites of HIV-1 Env vulnerability can be induced by infection, but simultaneous elicitation of neutralizing antibodies (NAbs) against multiple epitopes has not yet been achieved by vaccination. In this study, we designed a dual-epitope vaccine targeting both fusion peptide (FP) and V2 apex and evaluated its capacity to induce NAbs against both epitopes in rhesus macaques. This vaccine combined an FP conjugate with a cocktail of engineered Env trimers with enhanced V2 apex recognition and increased antigen retention in lymph nodes. Immunization of macaques with the dual-epitope vaccine elicited >1000-fold higher autologous tier 2-neutralization titers than the wildtype Env trimer and enhanced heterologous NAb breadth. Both FP and V2-apex monoclonal antibodies (mAb) were isolated from immunized macaques and showed heterologous neutralization with genetic and structural signatures that were similar to well-characterized FP and V2 apex bNAbs, although the V2 apex mAbs showed incomplete maturation. These results demonstrate proof-of-concept for simultaneous vaccine elicitation of NAbs against multiple sites of Env vulnerability, which will likely be critical for an effective HIV-1 vaccine. HIGHLIGHTSO_LIDesigned a dual-epitope vaccine targeting both fusion peptide (FP) and V2 apex C_LIO_LIV2-SET Env trimer conferred higher binding of V2 apex bNAbs, longer retention in draining lymph nodes, and >1000-fold higher induction of autologous neutralization titers compared with wildtype Env trimer C_LIO_LIDual-epitope vaccine enhanced serum tier 2 neutralization breadth C_LIO_LIDual-epitope vaccine elicited FP- and V2 apex-specific neutralizing mAbs with genetic signatures similar to well-characterized FP and V2 apex bNAbs C_LI

immunology↗

bioRxiv2025

Enterococcus faecalis delivers Obg GTPase via extracellular vesicles to instigate mTOR activity and promote HCC tumorigenesis

Hepatocellular carcinoma (HCC) is a malignant tumor that has been associated with dysbiosis of the gut microbiota. However, how the gut microbiota plays an oncogenic role in HCC remains largely unknown. Here, we show that Enterococcus faecalis (E. faecalis) is highly enriched in liver tumor tissues and is positively correlated with pathogenesis of HCC. E. faecalis promotes liver cancer cell proliferation, protein translation, cell migration and tumorigenesis. Mechanistically, we found that EF-derived extracellular vesicles (EF-EVs) deliver EF-Obg GTPase to activate host mTOR, thereby promoting liver cancer progression. Intriguingly, the EF-Obg protein exerts its influence on the mTOR pathway via a Ras-like G domain involved in GTP binding. EF-Obg exhibits a striking homology with the G1 site of the G domain of Rheb, a key positive regulator of mTOR. Obg gene is critical for E. faecalis to activate mTOR to promote hepatocarcinogenesis based on an engineered obg knockdown strain by CRISPR interference experiment. Clinically, abundant EF-Obg protein expression is correlated with enhanced activation of mTOR, leading to poor overall survival in HCC patients. Significantly, treatment of mTOR inhibitor Everolimus confers effectiveness in EF-colonized liver cancer orthotopic model, suggesting that Everolimus therapeutic approach can be effective for liver cancer patients with enriched E. faecalis. Taken together, we provide mechanistic and functional evidence to verify a direct causal relationship between tumor-resident E. faecalis enrichment and liver carcinogenesis, revealing that EF-Obg functions as a previously unidentified cross-kingdom activator of mTOR to promote liver tumorigenesis.

cell biology↗

bioRxiv2025

Cellular and Spatial Drivers of Unresolved Injury and Functional Decline in the Human Kidney

Building upon a foundational Human Kidney resource, we present a comprehensive multi-modal atlas that defines spatially resolved versus unresolved repair states and mechanisms in human kidney disease. Homeostatic interactions between injured kidney epithelium and its surrounding milieu determine successful repair outcomes, while pathogenic signaling promotes unresolved inflammation and fibrosis leading to chronic disease. We integrated multiple single-cell and spatial modalities across [~]700 samples from >350 patients ([~]250 research biopsies), analyzing [~]1.7 million cells alongside complementary mouse multi-omic profiles spanning acute-to-chronic injury and aging (>300,000 cells) and spatial transcriptomic analysis of >150 human biopsies. This cross-species atlas delineates functional pathways and druggable targets across the nephron and defines gene regulatory networks and chromatin landscapes governing tubular, fibroblast, and immune cell transitions from injury to either recovery or failed repair states. We identified distinct cellular states associated with specific pathological features that show dynamic distributions between acute kidney injury (AKI) and chronic kidney disease (CKD), organized within unique spatial niches that reveal progression mechanisms from early injury to unresolved disease. Gene regulatory analyses prioritized key transcription factor activities (SOX4, SOX9, NFKB1, REL, KLFs) and their target networks establishing disease states and tissue microenvironments. These regulatory programs were directly linked to clinical outcomes, identifying molecular signatures of recovery and secreted biomarkers predictive of AKI-to-CKD progression, providing a key resource for therapeutic development and precision medicine approaches in kidney disease.

molecular biology↗

bioRxiv2025

Pocket-based molecule generation with an SE(3)-equivariant language model leads to a potent and selective HPK1 inhibitor with in vivo efficacy

Deep learning shows promise in structure-based drug discovery, yet challenges persist in generating pharmacologically plausible molecules with valid 3D conformation and decent binding mode in the pocket. We introduce SE3-BiLingoMol, an SE(3)-equivariant Transformer for pocket-based 3D molecule generation, addressing two key limitations of existing language-model approaches. First, it uses Geometric Algebra Transformers for SE(3)-equivariant handling of continuous 3D coordinates. Second, a bidirectional attention mechanism mitigates conformational errors accumulated during autoregressive sampling. These innovations enable SE3-BiLingoMol to generate 2D drug-like, 3D geometrically valid molecules with superior binding modes. Validated on DUD-E dataset containing over 100 targets, the model achieved state-of-the-art performance in de novo design and optimization. We applied SE3-BiLingoMol to design potent and selective inhibitors for HPK1, a promising immunotherapy target. Through an iterative human-AI workflow, integrating AI generation with experimental validations (X-ray crystallography, bioassays), we identified Cmpd. 6. This novel tetracyclic compound demonstrates potent HPK1 inhibition, excellent cellular activity, favorable pharmacokinetics, and robust anti-tumor in vivo efficacy as monotherapy and with PD-1 blockade. Our work establishes a sophisticated generative AI framework for 3D molecule design and demonstrates its application in developing cancer immunotherapy. Highlights[bullet] SE(3)-Equivariant Language Model for Molecule Generation: We introduce a dual-channel SE(3)-equivariant language model that supports de novo molecule design and optimization, addressing key challenges in prior language model-based 3D molecule generation approaches. [bullet]Discovery of Potent HPK1 Inhibitors: Application of this model led to the discovery of highly selective HPK1 inhibitors with robust in vivo efficacy, resulting in a promising lead compound. [bullet]AI-Guided Rational Drug Development Paradigm: This work establishes a paradigm for AI-guided rational drug development, showcasing AI as a "co-pilot" that inspires drug designers with novel ideas in an "evidence-based context".

bioinformatics↗