Biology subjects

Xu, K.

Publications and source records attributed to Xu, K..

At least 19 recordsLinked to original sources

bioRxiv2026

Janusein and EB Cooperate to Enhance Microtubule Plus-end Catastrophe

Microtubule (MT) plus ends are organized by end-binding (EB) proteins, yet how EB-associated factors modulate the transition from growth to catastrophe remains incompletely understood. Using spatially resolved proximity proteomics, we identified Janusein (ZC2HC1A) as a conserved, poorly characterized MT-associated protein (MAP). At low expression levels, Janusein tracked MT plus ends and reduced EB comet velocity; at higher levels, it redistributed to the MT lattice and was associated with MT bundling and widespread MT loss. Loss of Janusein resulted in spindle defects in human cells. In vitro, Janusein alone facilitated MT elongation through its C-terminal fragment, but promoted MT catastrophe through its N-terminal fragment in the presence of EB proteins. Molecular dynamics simulations and mutational analyses suggest that an N-terminal loop may perturb lateral contacts between adjacent MT protofilaments. Together, these findings identify Janusein as an unconventional MAP that functionally cooperates with EB proteins to promote MT catastrophe.

cell biology↗

bioRxiv2026

Stressed alga-free coral cells can initiate coral bleaching

Coral bleaching is driving the rapid decline of coral reefs worldwide, yet a fundamental question remains unresolved: does the bleaching cascade begin in the coral host or in its intracellular algal symbionts, Symbiodiniaceae? Resolving where this process starts is critical for understanding how symbiosis collapses and where early intervention might be effective. Here we use transient stress to create a temporal window between the early cellular response and subsequent symbiosis breakdown. Across transient menthol and thermal stress in soft and stony corals, bleaching developed after the stress had ended. In Xenia sp., single-cell transcriptomics revealed a broad host response before detectable symbiont loss, including coordinated redox changes consistent with conditions favouring reactive oxygen species (ROS) accumulation. Cell-resolved imaging showed that alga-free coral cells were the first to become ROS-positive, whereas Symbiodiniaceae showed no detectable ROS signal. This early host response was transmissible. Conditioned medium from stressed cultures induced the emergence of ROS-positive host cells and subsequent symbiosis breakdown in unstressed recipients. Purified ROS-positive alga-free coral cells alone were sufficient to trigger symbiosis breakdown in recipient cultures without direct stress exposure. Suppressing the early emergence of ROS-positive coral cells markedly reduced later bleaching, whereas intervention after stress was substantially less effective. These findings identify a host-initiated, intercellularly propagating route to coral bleaching and reveal an early host response that can determine the subsequent fate of the symbiosis.

cell biology↗

bioRxiv2026

Evaluating Clustering and Imputation Approaches for Multivariate Longitudinal Clinical Data Using a Real-World eCDF-Copula Resampling Framework

The rapid growth of technologies such as wearable devices and high-throughput sequencing has generated unprecedented volumes of longitudinal data, often with multiple correlated features. This trend highlights the need for robust clustering methods tailored to multi-feature longitudinal datasets. Although previous reviews have benchmarked clustering methods in this context, most evaluations relied heavily on simulated and overly simplified datasets, which substantially limit their interpretability and applicability to real-world data. Missing data is also ubiquitous in longitudinal data, yet understudied in the context of clustering. To address these gaps, we systematically evaluate clustering methods across three distinct types of real-world longitudinal datasets: (1) a high-dimensional, low-timepoint metagenomic study, (2) a low-dimensional, high-timepoint multi-site repeated-measure study, and (3) a high-dimensional, high-timepoint repeated-measure study. We further investigate how missing data and different imputation strategies influence clustering performance. In addition, we introduce a novel resampling framework based on the empirical cumulative distribution function (eCDF) and a copula. This framework enables us to simulate multi-feature longitudinal data that preserves the correlation structure of real datasets, while also allowing for controlled variations such as sample size. Our simulation framework naturally accommodates missing data, thereby providing a flexible tool for benchmarking and comparing clustering methods in more realistic longitudinal settings.

bioinformatics↗

bioRxiv2026

m6A-Driven Intratumoral Cholesterol Biosynthesis Fuels Castration-Resistant Prostate Cancer Progression

Both nuclear pore complexes (NPCs) and RNA N6-methyladenosine (m6A) machinery are indispensable for proper cellular function. Although their collaborative roles in the nuclear export of messenger RNAs (mRNAs) have been reported, it remains ambiguous whether and how this collaboration may contribute to cancer progression. Here we identify a functional cooperation between NPCs and m6A signaling that promotes the development of castration-resistant prostate cancer (CRPC). We showed that nuclear export of m6A-modified mRNAs, mediated by the interaction between RNA methyltransferase METTL3 and the nucleoporin NUP93, is functionally coupled to cholesterol biosynthesis. Given that cholesterol-fueled intratumoral androgen production is one of the mechanisms driving CRPC, we demonstrated that overexpression of the wild-type METTL3 or NUP93, but neither the enzymatically dead METTL3 nor the mutant NUP93 that loses METTL3-interacting capability, elevates intracellular levels of androgens, activates AR signaling under castrate condition, and promotes androgen-independent growth of prostate cancer cells both in vitro and in vivo. Importantly, pharmacological inhibition of METTL3 or targeted demethylation on mRNAs encoding key cholesterol biosynthesis enzymes effectively suppressed CRPC malignancy. Together, these findings uncover a therapeutically targetable m6A-METTL3-NUP93 axis that links nuclear mRNA export and metabolic reprogramming to fuel CRPC progression, providing a conceptually new strategy for the treatment of this lethal disease.

cancer biology↗

bioRxiv2026

A prefrontal cortex-hypothalamus circuit for heart rate control in non-human primates

Humans exhibit changes in heart rate during cognitive and emotional events, and dysregulation of this brain-heart coupling is a hallmark of many psychiatric disorders. Yet, how circuits that regulate cognition exert direct control over heart physiology remains poorly understood, particularly in primates. Here, we identify a prefrontal-hypothalamic circuit that exerts powerful, bidirectional control over heart rate in monkeys. We show that electrical stimulation of a small ventral subregion within the ventrolateral prefrontal cortex (area 47/12; a12) evokes rapid and robust decreases in heart rate and increases heart rate variability, whereas neighboring prefrontal regions produce weaker or no effects. And a12 neurons are coupled to heart rate on a moment-by-moment basis. Anatomical tracing data reveal projections from a12 to a circumscribed region of lateral hypothalamus (LHA), and stimulation of this LHA target recapitulates the cardiac effects of prefrontal activation. Consistent with a prefrontal-to-hypothalamus system for heart rate control, stimulation of LHA produces faster, stronger, state-independent heart effects, whereas stimulation of a12 produces effects that are sensitive to internal states. Oppositely to its activation, chemogenetic inactivation of a12 or its projections to LHA increases heart rate, demonstrating that this pathway can provide tonic inhibitory control over cardiac function. To make steps towards a more cell-type-specific and minimally invasive causal control of this circuit, we developed a primate sonogenetic approach based on ultrasound activation of CaMKII+ neurons virally transduced with TRPV1. Sonogenetic stimulation of a12 produced increases in heart rate variability, and these effects were attenuated by inactivation of LHA. Finally, we found that sonogenetic and chemogenetic perturbation of the same prefrontal region during a value-based decision-making task bidirectionally alters risky decision making. Together, our work identifies how a12 regulates heart rate and cognition, revealing a mechanistic substrate for brain-body interactions and a potential therapeutic target for disorders of cognitive-autonomic dysregulation.

neuroscience↗

bioRxiv2026

Transcription factors read a second regulatory code in chromatin

Transcription factors (TFs) decode gene regulatory information written in DNA, yet how this vocabulary is interpreted within chromatin remains largely unexplored. Here, using an upgraded NCAP-SELEX platform, we systematically map the nucleosomal DNA recognition landscapes of 269 human TFs. We uncover a widespread, chromatin-dependent mode of sequence recognition: many TFs recognize motifs on nucleosomal DNA that are distinct from their canonical naked-DNA binding sites, revealing that nucleosome architecture encodes a second gene regulatory code in chromatin. Cryo-electron microscopy structures of TF-nucleosome complexes demonstrate that this second code is read through a combination of nucleosome-induced DNA deformation and direct protein-histone contacts. Functional analyses show that these chromatin-encoded motifs actively promote chromatin accessibility and drive cell-type-specific cis-regulatory activity in vivo. Strikingly, a single TF can deploy its canonical and nucleosome-derived motif repertoires to partition and govern entirely distinct physiological programs. Together, our findings establish that TFs interpret two complementary layers of genomic information: the primary DNA sequence and a second code written into the nucleosome architecture. This chromatin-encoded layer of regulatory information fundamentally expands our understanding of how TF specificity is achieved and how gene regulatory networks are wired in multicellular organisms.

systems biology↗

bioRxiv2026

Genetic drift decouples Fisherian trait-preference coevolution from genetic correlations in finite populations

A central mechanism of sexual selection theory is Fisher's process, which refers to the coevolution of male traits and female preferences, in which preference is indirectly selected through genetic association with male trait alleles built up via mate choice. However, empirical studies often fail to detect strong trait-preference genetic correlations, raising doubts about the importance of Fisherian selection in nature. Notably, the theoretical expectation that genetic correlations are essential for trait-preference coevolution through Fisherian selection derives largely from models assuming infinitely large populations, whereas real populations are finite. Using population genetic models, I show that interactions between Fisherian selection and genetic drift can fundamentally decouple trait-preference genetic correlations from the evolution of male traits and female preferences. Genetic drift generally reduces expected trait-preference correlations but simultaneously promotes the expected increase in female preference frequency beyond deterministic predictions. Consequently, in populations of realistic sizes, trait-preference correlations may often be weak or even negative, particularly when recombination among trait and preference loci is infrequent, but substantial trait-preference coevolution can still occur. Therefore, the strength of genetic correlations may be an unreliable indicator of the extent of trait and preference coevolution, offering a potential resolution to a longstanding dilemma in sexual selection theory.

evolutionary biology↗

bioRxiv2026

The probability of evolutionary rescue with competition: scaling-up from a single species to a community

When ecological communities experience environmental change, continued coexistence may require rapid adaptation to avoid extinction, i.e., evolutionary rescue. This rescue can be inhibited by competition. Combining Lotka-Volterra and population genetic models, we investigate how both within- and between-species competition impact the probability of rescue. With a single species we derive closed-form approximations that quantify the negative impact of competition. We then use a two-species model to examine how the sweep of a rescue mutant in one species inhibits establishment of rescue mutants in another, a process we term survival interference. This interference is strongest when the stronger competitor begins recovering first, creating eco-evolutionary priority effects despite conditions permitting deterministic coexistence. Finally, considering larger competitive communities, we document how survival interference reduces the number of rescued species but weakens with species richness and unevenness. These results move us closer to predicting evolutionary rescue in ecological communities.

ecology↗

bioRxiv2026

Eco-evolutionary feedbacks generate bistability in population persistence under gradual environmental change

Understanding how populations persist in gradually deteriorating environments through evolution is a central question in ecology and evolutionary biology. Previous studies have primarily focused on identifying the critical rate of environmental change beyond which extinction is certain. However, the existence of a viable equilibrium when the rate is below the threshold does not guarantee that a population can survive the transient dynamics to reach it. Using a quantitative genetic model that explicitly incorporates feedback among population size, genetic variance, and mean trait evolution, we show that population persistence can exhibit bistability when the rate of environmental change is below the extinction threshold. Specifically, extinction still occurs if the initial population size and genetic variance fall below a critical level. The initial state also influences the eco-evolutionary dynamics, such that a temporary increase or decline in population size and/or genetic variance does not necessarily predict the ultimate fate of the population. Therefore, in addition to estimating the critical rate of environmental change for extinction, characterizing current population size, genetic variation, and the degree of maladaptation may improve predictions of extinction risk in deteriorating environments.

evolutionary biology↗

bioRxiv2026

In vitro and in silico characterization of competitive inhibition and repression of DUX4 target gene activation as a therapeutic approach for facioscapulohumeral muscular dystrophy (FSHD)

Facioscapulohumeral muscular dystrophy (FSHD) is a rare neuromuscular disease caused by aberrant re-expression of the embryonic transcription factor DUX4 in skeletal muscle, which activates a toxic transcriptional program that drives progressive muscle wasting. No disease-modifying therapies currently exist. Prior work in mammalian and zebrafish models has shown that a truncated form of DUX4 retaining only its DNA-binding domain (DBD) lacks transactivation capacity and can suppress DUX4-FL-driven pathology; separately, dCas9/KRAB-based epigenetic repressors have demonstrated efficacy in silencing DUX4 expression, though CRISPR-based strategies face challenges from the repetitive nature of the D4Z4 locus, the immunogenicity associated with bacterial Cas proteins, and the payload limitations of gene delivery vehicles. Building on these findings, we corroborate that the DUX4 DBD, comprising both homeodomains, acts as a non-toxic competitive inhibitor of full-length DUX4 (DUX4-FL) at its genomic target sites, and extend this strategy by fusing the DBD to a human KRAB(ZNF10) domain, converting DUX4 from a transcriptional activator into a fully humanized epigenetic silencer of its own targets. Using a fluorescent DUX4-responsive reporter, we show that DBD alone produces dose-dependent repression of DUX4-FL transcriptional activity in HEK293T cells (200-fold at the highest inducible dose tested), while a constitutively expressed DBD-KRAB fusion produces significantly greater repression than DBD alone (949-fold versus 17-fold at a 25x molar ratio), with a similar trend observed in C2C12 myoblasts (47-fold versus 3.3-fold knockdown). To contextualize these findings and explore dosing considerations, we developed three complementary computational models - a transcription factor competitive binding model, a myotube diffusion model, and an ordinary differential equation (ODE) compartmental model - that illustrate how DBD concentration, intracellular diffusion, and population-level cell state transitions may relate to therapeutic efficacy. Together, these results corroborate and extend existing approaches into a single, fully humanized construct that may help circumvent the immunogenicity and delivery limitations of Cas-based systems.

bioengineering↗

bioRxiv2026

A hierarchical activation threshold shapes RSV F antibody repertoires from toddler vulnerability to adult protection

Respiratory syncytial virus (RSV) causes severe infant morbidity, yet the mechanisms underlying their suboptimal immunity and the failure of prefusion F-stabilized adult vaccines in infants remain unclear. Here we integrate deep mutational scanning of 761 preF-binding antibodies, repertoire profiling of 102 repeatedly exposed pediatricians and 61 RSV-experienced toddlers, structural analysis, and >40,000 viral genomes to decode RSV F immunity. We resolved 12 immunologically distinct functional antibody subclasses and uncovered a hierarchical activation threshold: apical epitope-targeting antibodies with superior neutralizing potency demand extensive somatic hypermutation and cooperative CDR networks, whereas central-to-basal epitope-directed, marginal or non-neutralizing antibodies engage germline-encoded antibodies through minimal mutations (e.g., S31G). Toddler repertoires are confined to low-threshold, non-productive sites; adult repertoires enrich for apical elite neutralizers whose immune pressure drives contemporary RSV-B evolution. These reframe pediatric RSV vulnerability as a threshold-gated repertoire deficit and prescribe vaccine strategies that actively redirect immunodominance from permissive epitopes toward high-barrier apical targets.

immunology↗

bioRxiv2026

Human antibody heavy chain variable region exon interacts with 3' regulatory region to form a stable somatic hypermutation center

Activation-induced cytidine deaminase (AID) initiates somatic hypermutation (SHM) of immunoglobulin heavy chain (HC) variable region exons in germinal center (GC) B cells, allowing selection of affinity-matured B cell receptors. The V(D)J exon location is privileged for SHM targeting of diverse sequences within it. In analogy to the related HC class switch recombination center, we proposed that cohesin-mediated loop extrusion juxtaposes widely separated enhancers, including 3 Igh regulatory region ("3RR"), with the V(D)J exon in GC B cells to establish a SHM-center (SHM-C) "privileged" for AID access. To test this hypothesis, we employed a human RAMOS GC B lymphoma cell SHM model with inducible AID expression in which we targeted an unmutated mouse HC V(D)J exon in place of its human counterpart. Activation of AID expression in this cycling RAMOS line for 10 days led to robust SHM accumulation in the mouse V(D)J exon with a pattern similar to that in mouse GC B cells. In this model, the mouse V(D)J exon was indeed juxtaposed to the two human 3RRs and other elements consistent with SHM-C formation. While deletion of individual 3RRs had little effect, deletion of both greatly reduced SHM, but left V(D)J exon transcription unabated. Moreover, the SHM-C and SHM accumulation remained intact over a 10-day period in viably G1-arrested and viably G1-arrested RAD21 (cohesin)-depleted versions of the model. We discuss implications for SHM-C structure and function, its long-term functional stability in the absence of the loop extrusion process proposed to assemble it, and for 3RR SHM functions beyond transcriptional activation. Significance StatementUpon encountering pathogens, B cells enter specialized lymphoid structures known as germinal centers (GCs). GC B cells express activation-induced cytidine deaminase (AID) that initiates somatic hypermutation (SHM) of the portion of antibody genes that encodes antigen recognition. B cells in which mutations increase antibody affinity for activating pathogen are selected to enhance antibody responses. We proposed SHM occurs in a chromosomal structure in the antibody gene locus termed a SHM-center (SHM-C). By studying SHM in a human B cell lymphoma model, we provide direct evidence for existence of a SHM-C assembled from different functional components normally separated by considerable distances. We also find that the SHM-C is surprisingly stable and continues to function for many days in non-dividing cells.

immunology↗

bioRxiv2026

Bacterial surface display enables lysis-independent joint host-pathogen single-cell profiling

Heterogeneity in host-pathogen interactions arises from variation in both host cell state and pathogen state, yet most single-cell methods capture these features separately. Joint profiling is limited by fundamental technical mismatches between host- and pathogen-derived material, particularly differences in cell wall structure, lysis requirements, and molecular abundance. Here we introduce a lysis-independent strategy that enables unified measurement of intracellular bacterial presence and state alongside host single-cell profiles. We repurpose bacterial surface display to encode promoter activity and bacterial identity as antibody-detectable signals, rendering bacterial features compatible with existing antibody-based single-cell assays. This approach is modular across promoters, epitope tags, display scaffolds, and bacterial species, including Escherichia coli and Mycobacterium tuberculosis. Surface-displayed reporters are detectable during intracellular infection and can be read out by flow cytometry and droplet-based single-cell RNA sequencing without pathogen-specific lysis optimization or pre-sorting on pathogen signal. Applying this method to infected macrophages, we link heterogeneous bacterial uptake to heterogeneous expression of phagocytosis-associated host programs. This strategy enables scalable, joint host-pathogen single-cell measurements and expands the range of pathogens and states accessible to high-throughput single-cell analysis.

bioengineering↗

bioRxiv2026

Quantifying Cross-Modal Shared Information Between Histomorphology and Spatial Transcriptomics via Spatiotemporal Trajectory Correlation

Histopathological imaging and spatial transcriptomics (ST) provide synergistic morphological and molecular insights into tissue architecture. While conventional downstream analyses predominantly adopt a discrete paradigm, such as segmenting histological images or identifying spatial domains, emerging trajectory reconstruction methods offer a continuous perspective for analyzing these modalities. However, most current studies are confined to single-modality or single-organ analyses, lacking systematic integration across modalities and multiple organs. To address this limitation, we performed trajectory reconstruction on ST-derived gene expression data and on histopathological morphological features extracted using ten widely used pathology pretrained models across multiple cancer samples from six organs. The results demonstrate that trajectory reconstruction, as a continuous analytical framework, effectively bridges spatial transcriptomics and histopathological imaging. More importantly, we propose an innovative framework that uses trajectory pseudotime as a mediating variable to quantify the extent of information sharing between molecular and morphological features. This framework not only provides a new perspective for understanding the intrinsic links between modalities but also establishes a solid theoretical and methodological foundation for future cross-modal translation studies.

pathology↗

bioRxiv2026

Protein-based, live-cell PAINT microscopy with fluorogenic exchangeable HaloTag ligands

We report a generalizable method for single-molecule localization super-resolution microscopy in living cells. Live-cell PAINT (points accumulation for imaging in nanoscale topography) and single-molecule diffusivity mapping (SMdM) of intracellular protein targets are achieved by pairing exceptionally fluorogenic bis-trifluoromethyl rhodamine (BF) dyes with reversible HaloTag ligands and widely available HaloTag fusion proteins. This far-red small-molecule update to protein-based PAINT is readily incorporated into existing super-resolution microscopy workflows: wash-free staining and imaging are achieved for cell lines and primary neuron cultures for diverse protein targets. Pairing with photoconvertible fluorescent proteins further enables simultaneous two-color live-cell super-resolution microscopy and SMdM.

cell biology↗

bioRxiv2026

PFOS aggravates atherosclerosis via Bacteroides caecimuris expansion-driven bile acid remodeling and subsequent intestinal FXR-TLR3 signaling cascade

BACKGROUNDPerfluorooctane sulfonate (PFOS) is a widely distributed persistent organic pollutant in the environment and has been associated with an increased risk of atherosclerosis. However, the underlying pathogenic mechanisms remain largely unclear. This study aimed to investigate the effects of PFOS on atherosclerosis and its associated gut-vascular axis. METHODSPseudo-germ-free mouse models and fecal microbiota transplantation (FMT) were used to determine the role of the gut microbiota in PFOS-induced atherosclerosis. Metagenomic sequencing was performed to characterize alterations in gut microbial composition following PFOS exposure, and targeted metabolomics was used to assess bile acid profiles in the ileum and plasma. Transcriptomic analysis of Bacteroides caecimuris (B.caecimuris) was conducted to explore the reasons for the increased abundance of B.caecimuris after PFOS exposure. In addition, intestinal transcriptomics and ChIP-qPCR were performed to validate transcriptional regulation within the FXR-TLR3 signaling axis. RESULTSAmong 127 participants with paired serum and fecal samples, including 82 patients undergoing coronary angiography with Gensini scores (GS score), fecal PFOS levels were significantly associated with lipid profiles and GS score, whereas serum PFOS showed no such association.Mechanistically, PFOS exposure promotes intestinal enrichment of B.caecimuris by upregulating its tolC gene, thereby enhancing efflux capacity.This microbial shift was accompanied by reduced levels of tauro-ursodeoxycholic acid (TUDCA) and aberrant activation of intestinal FXR signaling.Further analyses demonstrated that FXR activation upregulated TLR3 expression and promoted inflammatory responses and atherosclerosis progression via the TLR3-NF-{kappa}B signaling axis. Both intestinal epithelial-specific FXR deficiency (Fxr{Delta}IE) and TUDCA supplementation significantly suppressed pathway activation and alleviated disease phenotypes.Functional experiments identified TLR3 as a key downstream effector of FXR. Overexpression of TLR3 abolished the protective effects observed in Fxr{Delta}IE mice. Moreover, pharmacological inhibition of TLR3 using CU CPT-4a significantly improved established atherosclerotic lesions in vivo. CONCLUSIONSThis study identifies a gut microbiota-driven FXR-TLR3 signaling axis that mediates PFOS-induced atherosclerosis. These findings provide new mechanistic insights into environmentally induced cardiovascular disease and suggest potential targets for risk assessment and therapeutic intervention.

pathology↗

bioRxiv2026

The membrane distal domain of CD16a allosterically regulates NK cell ADCC

Antibody-dependent cellular cytotoxicity (ADCC) by natural killer (NK) cells is mediated by the activating IgG receptor CD16a (Fc{gamma}RIIIa), yet the molecular mechanisms governing receptor activation remain poorly understood. We demonstrate that the membrane-distal domain 1 (D1) of CD16a functions as an allosteric checkpoint that controls ADCC independently of IgG-Fc binding. A nanobody, C28, that binds an electronegative patch in D1 dose-dependently blocks NK cell ADCC against multiple therapeutic antibodies without affecting direct cytotoxicity. A second nanobody, C21, binding an adjacent D1 epitope has no such effect. Cryo-EM structures of the CD16a-IgG-nanobody complex reveal that C28 allosterically competes with core-fucosylated IgG and stabilizes a closed D1 conformation resembling unliganded receptor, even when Fc is bound. Molecular dynamics simulations show that occupation of the D1 epitope rigidifies the IgG-binding site, stabilizing CD16a overall in contrast with IgG binding alone. The nanobody C28 restricts CD3{zeta} phosphorylation in both resting and ADCC-activated NK cells, revealing tonic inhibitory control upstream of the signaling cascade. Using MINFLUX nanoscopy, we also show that CD16a forms dimers of [~]9 nm spacing on the NK cell surface, a geometry unaltered by the ADCC-enhancing L48H polymorphism. Drawing on structural parallels with the IgE receptor Fc{varepsilon}RI, which is held inactive as a cholesterol-stabilized dimer, we propose that CD16a dimerization through D1 contacts represents a conserved autoinhibitory mechanism among Fc receptors. Consistent with this model, structure-guided disruption of the C28 epitope in NK-92 cells enhances ADCC potency and killing kinetics, providing a blueprint for engineering improved cellular immunotherapeutics.

immunology↗

bioRxiv2026

StructureSAFE: A structure-aware chemical language model for unified hit identification and lead optimization

Structure-based generative models (SBGMs) hold great promises for accelerating drug discovery by enabling target-aware molecular design. However, existing approaches face fundamental challenges: three-dimensional graph-based models can explicitly incorporate protein structural information but often generate chemically implausible molecules due to limited training data, while chemical language models (CLMs) produce chemically plausible molecules but struggle to effectively leverage three-dimensional structural information for structure-conditioned generation and hard to incorporate lead optimization functionality due to the nature of SMILES string. Here, we present StructureSAFE, a structure-aware chemical language model that resolves this trade-off by integrating protein structural and evolutionary encoders with the SAFE molecular representation via pretraining and finetuning training scheme, enabling both de novo hit identification and a comprehensive suite of lead optimization subtasks within a unified framework. Comprehensive benchmarking on the MolGenBench dataset demonstrates that StructureSAFE achieves state-of-the-art (SOTA) performance across multiple metrics, with particularly pronounced improvements in chemical plausibility relative to graph-based models lacking pretraining. Evaluation on a rigorously constructed held-out test set further confirms its ability to generate drug-like, synthetically accessible molecules with competitive predicted binding affinities for previously unseen targets on both hit identification and lead optimization setting. In silico case studies across four therapeutically relevant targets validate its capacity to generate chemically plausible molecules that recapitulate key binding interactions of known high-affinity ligands while proposing novel interactions for potential better affinity and exploring previously unknown regions of chemical space. Taking together, StructureSAFE represents a versatile and practical tool to provide high-quality candidate molecules for augmenting medicinal chemistry workflows in both hit identification and lead optimization campaigns.

bioinformatics↗