Biology subjects

Kim, W. Y.

Publications and source records attributed to Kim, W. Y..

14 recordsLinked to original sources

bioRxiv2026

Predicting conformational changes induced by biomolecular interactions

Biomolecular interactions trigger structural changes of protein complexes essential for cellular signaling, regulation, and drug action. Modern co-folding models can now generate detailed structures of multi-component complexes directly from sequence. However, recent evaluations reveal that they frequently fail to capture specific conformational states induced by a ligand or partner. Here, we show that explicitly modeling the structural transition from an unbound to a bound state, rather than generating the bound complex alone, better resolves these interaction-induced states. Our generative learning framework, K-Fold, transforms independently predicted unbound component priors into an assembled complex through an ultrafast diffusion bridge, operating entirely without multiple sequence alignments. By eliminating co-evolutionary dependencies, K-Fold accurately recovers non-coevolved interfaces in antibody-antigen and targeted-degrader complexes, as well as ligand-directed regulatory states in GPCRs and kinases. Furthermore, by coupling this diffusion bridge with diverse monomeric equilibrium ensembles, K-Fold provides a computationally tractable foundation for sampling complex multi-state landscapes without retraining. These results demonstrate that formulating molecular assembly as an explicit structural transition moves structure prediction beyond static geometries toward capturing biomolecular dynamics.

molecular biology↗

bioRxiv2026

Synthesis-aware generative design in trillion-scale chemical spaces for automated drug discovery

Autonomous drug discovery via generative molecular design is critically bottlenecked by the production of chemically intractable structures. Multi-trillion-scale make-on-demand libraries guarantee synthetic feasibility, but conventional virtual screening cannot efficiently navigate these vast spaces or address multi-parameter optimization. We introduce Hyper Screening X, a generative AI framework integrating structure-based design with automated synthesis hardware. By encoding deterministic reaction logic in a generative flow network, the framework evaluates only 10 million candidates to navigate an 11-trillion-compound space, jointly optimizing physicochemical properties while ensuring compatibility with automated synthesis. Prospective validation against an SLC1A5 variant with a cryptic interface yielded a 96% synthesis success rate and a 50% functional hit rate, identifying two first-in-class lead compounds. This synthesis-aware strategy advances autonomous, closed-loop drug discovery.

bioinformatics↗

bioRxiv2026

How do Co-Folding Models Organize Structural Information?

Co-folding models accurately predict biomolecular complexes, but how their internal representations support joint structure prediction remains unclear. We analyze the Boltz-1 trunk by decomposing its pair representation into intra-chain and inter-chain blocks and applying ablation, layer-wise representation geometry, and activation patching. Our results show that structural information is distributed across three streams: the single representation carries token-internal atomic detail, the intra-chain pair representation encodes chain-level geometry, and the inter-chain pair representation controls relative molecular arrangement. Across the trunk depth, representations follow a Mix-Compress-Refine trajectory, while linear probes reveal that intra-chain distance content is largely supplied by MSA preconditioning and inter-chain distance content is constructed progressively across Pairformer layers at each recycle. Apo-holo patching further shows that ligand-induced backbone changes are not fully resolved before diffusion; instead, the diffusion module reconciles partially inconsistent intra- and inter-chain constraints. These findings provide a mechanistic view of how co-folding trunks organize structural information and suggest more partition-aware designs for future biomolecular structure prediction models.

bioinformatics↗

bioRxiv2026

AtlasFold: Protein structure prediction with metagenomic-scale language models

Protein language models (PLMs) trained on evolutionary sequences learn representations that encode protein structure, enabling direct structure prediction without multiple-sequence alignments (MSAs). Here we present the Atlas model family, an open and trainable system spanning protein language modeling, monomer folding, and protein-complex prediction. AtlasLM-3B is a 3B-scale language model trained with masked language modeling on approximately 1.56 billion sequences, including metagenomic data, and outperforms the similarly sized ESM2-3B in unsupervised contact prediction. Building on these representations, AtlasFold predicts all-atom protein structures and achieves state-of-the-art accuracy among PLM-based folding models. Fine-tuning AtlasFold for protein-complex prediction produces AtlasFold-Multimer (AtlasFold-M), whose antibody-antigen prediction performance is comparable to that of AlphaFold3 and ESMFold2. This protein-specific folding architecture enables fast, memory-efficient inference with AtlasFold and AtlasFold-M. By releasing the training code and data, stage checkpoints, and model weights under the MIT License, we provide a foundation for advancing PLM-based protein structure prediction.

bioinformatics↗

bioRxiv2026

Vertical profile of airborne microbial communities in the Southern Ocean atmospheric boundary layer

Aerobiological studies have largely focused on near-surface sampling and horizontal biogeographic patterns, while vertical structuring of airborne microbial communities within the atmospheric boundary layer (ABL) remains poorly understood. Here, we investigated microbial communities across the lower and upper ABL in a low-orography coastal site on the Antarctic Peninsula, representative of the Southern Ocean marine ABL and with low direct human influence. Airborne microorganisms were sampled simultaneously using ground-based and aerial platforms on five occasions. Community composition, abundance, and cell morphometry were assessed using metabarcoding and epifluorescence microscopy and interpreted alongside atmospheric observations. Airborne bacterial and eukaryotic communities showed consistent vertical stratification, although partial taxonomic overlap indicates vertical connectivity between atmospheric layers. Lower ABL communities were more diverse than upper ABL counterpart, compositionally homogeneous, and dominated by marine-associated taxa, reflecting strong influence from local sources and turbulent mixing. In contrast, upper ABL communities were less diverse but more heterogeneous among sampling events, enriched in stress-tolerant, terrestrial and plant-associated taxa, consistent with atmospheric filtering, selective upward transport, and long-range atmospheric inputs. Upper-layer samples also exhibited higher microbial abundance and greater prevalence of elongated cell morphologies, suggesting particle accumulation aloft and aerodynamic selection permanence. Together, these findings identify the Southern Ocean ABL as a vertically structured microbial habitat organized into two partially decoupled sublayers, in which atmospheric dynamics regulate microbial dispersal, ecosystem connectivity, and biogeographic patterns.

ecology↗

bioRxiv2026

Elucidating enzyme-substrate specificity through co-folding foundation model

Enzymatic catalysis relies on precise structural and chemical complementarity, yet systematically mapping enzyme-substrate interactions remains a critical bottleneck. While structure-aware methods have advanced functional annotation, their reliance on predefined binding pockets and rigid-body docking fails to capture the ligand-induced conformational changes essential for catalytic turnover. Here we introduce Boltz2ESI, an end-to-end framework that predicts enzyme-substrate interactions by leveraging structural knowledge learned by a biomolecular foundation model. Through native co-folding, the framework inherently captures active-site plasticity without requiring predefined pocket annotations. Integrating these learned biophysical priors with global evolutionary context and geometric molecular descriptors, Boltz2ESI consistently outperforms state-of-the-art sequence-based and rigid-docking approaches. Extensive validation demonstrates that the framework accurately discriminates tight sub-family specificities, enabling effective candidate prioritization for biosynthetic pathway elucidation, as demonstrated on the withanolide pathway. Ultimately, this structure-dynamic approach establishes an actionable foundation for accelerating rational biocatalyst discovery and large-scale pathway de-orphaning.

bioinformatics↗

bioRxiv2026

Deep Learning-Driven Discovery of Mitochondrial Factors Modulating Influenza A Virus Infection

Influenza A virus exploits host cellular machinery across subcellular compartments, yet the organelle-level changes that distinguish infected from uninfected cells and the molecular players driving them remain poorly defined. Here, we combine organelle image-based deep learning with proximity labeling chemoproteomics to address this gap. A convolutional neural network identified mitochondrial morphology as the strongest single-cell predictor of infection status (precision = 84.9%). Proximity labeling profiling of mitochondrial matrix proteome revealed 99 proteins with significantly altered upon infection, of which five (CH60, ETHE1, LONM, MPPB, and SQOR) were validated as host restriction factors whose depletion elevated interferon-{beta} expression, enhanced viral RNA accumulation, or increased progeny virus production. Notably, two of these factors, ETHE1 and SQOR, operate within the mitochondrial hydrogen sulfide oxidation pathway, and pharmacological scavenging of H2S by hydroxocobalamin dose-dependently reversed their knockdown phenotypes, directly linking mitochondrial sulfide metabolism to antiviral defense against influenza.

microbiology↗

bioRxiv2025

Aortic annuloplasty FSI digital twin of 3D-printed phantoms with 4D-flow MRI comparison

BackgroundAortic annuloplasty, involving the implantation of an external ring around the aortic root to reduce annular dimensions, is a promising treatment for aortic valve insufficiency. However, its hemodynamic effects remain underexplored due to the absence of computational models validated by experimental and clinical data. MethodsThis study introduces a computational fluid-structure interaction (FSI) model of supra valvular aortic annuloplasty using 4D-flow magnetic resonance imaging (MRI). Native and post-annuloplasty conditions of idealized aortic root phantoms, including the aortic valve, were CAD-modelled and 3D-printed with elastic resin. These phantoms were tested in a mock circulatory flow-loop providing normal pulsatile physiologic conditions using a glycerol-water mixture to simulate blood viscosity. Flow and pressure data collected from sensors were used as boundary conditions for FSI simulations. Experimental velocity fields from 4D-flow MRI were compared to computational results to assess model accuracy. ResultsMRI scans of the annuloplasty model showed an increased peak systolic velocity (up to 145.4 cm/s) and localized flow alterations, corresponding to a higher pressure gradient across the valve. During regurgitation, the annuloplasty model showed broader velocity distributions compared to the native condition. The FSI simulations closely matched 4D-flow MRI data, with strong correlation coefficients (r > 0.93) and minimal Bland-Altman differences, particularly during systolic phases. ConclusionsThis study establishes an integrative methodology combining in-vitro, in-silico, and clinical imaging techniques to evaluate aortic annuloplasty hemodynamics. The validated digital twin framework offers a pathway for patient-specific modelling, enabling prediction of surgical outcomes and optimization of aortic valve repair strategies.

bioengineering↗

bioRxiv2024

HOS15-mediated turnover of PRR7 enhances freezing tolerance

Arabidopsis PSEUDO RESPONSE REGULATOR7 (PRR7) is a core component of the circadian oscillator which also plays a crucial role in freezing tolerance. PRR7 undergoes proteasome-dependent degradation to discretely phase maximal expression in early evening. While its transcriptional repressive activity on downstream genes is integral to cold regulation, the mechanism of the conditional regulation of the PRR7 protein activity is unknown. We used double mutant analysis, protein interaction and ubiquitylation assays to establish that the ubiquitin ligase adaptor, HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENE 15 (HOS15), controls the protein accumulation pattern of PRR7 through direct protein-protein interactions. Freezing tolerance and electrolyte leakage assays show that PRR7 enhances cold temperature sensitivity, supported by ChIP-qPCR at C-REPEAT BINDING FACTOR (CBF) and COLD REGULATED 15A (COR15A) promoters where PRR7 levels were higher in hos15 mutants. We establish that HOS15 mediates PRR7 protein turnover through enhanced ubiquitylation at low temperature in the dark. Under the same conditions, increased PRR7 association with the promoter regions of CBFs and COR15A in hos15 correlates with decreased CBF1 and COR15A transcription and enhanced freezing sensitivity. We propose a novel mechanism whereby HOS15-mediated regulation of PRR7 provides an intersection between the circadian system and other cold acclimation pathways leading to freezing tolerance through upregulation of CBF1 and COR15A.

plant biology↗

bioRxiv2024

Determination of permissive and restraining cancer-associated fibroblast (DeCAF) subtypes

Cancer-associated fibroblast (CAF) subpopulations in pancreatic ductal adenocarcinoma (PDAC) have been identified using single-cell RNA sequencing (scRNAseq) with divergent characteristics, but their clinical relevance remains unclear. We translate scRNAseq-derived CAF cell-subpopulation-specific marker genes to bulk RNAseq data, and develop a single- sample classifier, DeCAF, for the classification of clinically restraining and permissive CAF subtypes. We validate DeCAF in 19 independent bulk transcriptomic datasets across four tumor types (PDAC, mesothelioma, bladder and renal cell carcinoma). DeCAF subtypes have distinct histology features, immune landscapes, and are prognostic and predict response to therapy across cancer types. We demonstrate that DeCAF is clinically replicable and robust for the classification of CAF subtypes in patients for multiple tumor types, providing a better framework for the future development and translation of therapies against permissive CAF subtypes and preservation of restraining CAF subtypes. SignificanceWe introduce a replicable and robust classifier, DeCAF, that delineates the significance of the role of permissive and restraining CAF subtypes in cancer patients. DeCAF is clinically tractable, prognostic and predictive of treatment response in multiple cancer types and lays the translational groundwork for the preclinical and clinical development of CAF subtype specific therapies.

genomics↗

bioRxiv2021

ZHX2 Promotes HIF1α Oncogenic Signaling in Triple-Negative Breast Cancer

Triple-negative breast cancer (TNBC) is an aggressive and highly lethal disease, which warrants the critical need to identify new therapeutic targets. We show that Zinc Fingers And Homeoboxes 2 (ZHX2) is amplified or overexpressed in TNBC cell lines and patients. Functionally, depletion of ZHX2 inhibited TNBC cell growth and invasion in vitro, orthotopic tumor growth and spontaneous lung metastasis in vivo. Mechanistically, ZHX2 bound with hypoxia inducible factor (HIF) family members and positively regulated HIF1 activity in TNBC. Integrated ChIP-Seq and gene expression profiling demonstrated that ZHX2 co-occupied with HIF1 on transcriptionally active promoters marked by H3K4me3 and H3K27ac, thereby promoting gene expression. Furthermore, multiple residues (R491, R581 and R674) on ZHX2 are important in regulating its phenotype, which correspond with their roles on controlling HIF1 activity in TNBC cells. These studies establish that ZHX2 activates oncogenic HIF1 signaling, therefore serving as a potential therapeutic target for TNBC.

cancer biology↗

bioRxiv2021

Pparg drives luminal differentiation and luminal tumor formation in the urothelium

Pparg, a nuclear receptor, is downregulated in basal subtype bladder cancers that tend to be muscle invasive and amplified in luminal subtype bladder cancers that tend to be non-muscle invasive. Bladder cancers derive from the urothelium, one of the most quiescent epithelia in the body which is composed of basal, intermediate, and superficial cells. We find that expression of an activated form of Pparg (VP16;Pparg) in basal progenitors induces formation of superficial cells in situ, that exit the cell cycle, and do not form tumors. Expression in basal progenitors that have been activated by mild injury however, results in luminal tumor formation. We find that tumors are immune deserted, which may be linked to downregulation of Nf-kb, a Pparg target. Interestingly, some luminal tumors begin to shift to basal subtype tumors with time, downregulating Pparg and other luminal markers. Our findings have important implications for treatment and diagnosis bladder cancer.

cancer biology↗

bioRxiv2021

α1-COP delivers sphingolipid modifiers and controls plasmodesmal callose deposition in Arabidopsis

Callose is a plant cell wall polymer in the form of {beta}-1,3-glucan, which regulates symplasmic channel size at plasmodesmata (PD). It plays a crucial role in a variety of processes in plants through the regulation of intercelluar symplasmic continuity. However, how to maintain callose homeostasis at PD in the molecular levels is poorly understood. To further elucidate the mechanism of PD callose homeostasis, we screened and identified an Arabidopsis mutant plant that exhibited excessive callose deposition at PD. Based on the Next-generation sequencing (NGS)-based mapping, other mutant allele analysis, and complementation assay, the mutated gene was shown to be 1-COP, which encodes a member of the COPI coatomer complex comprised of , {beta}, {beta}', {gamma}, {delta}, {varepsilon}, and {zeta} subunits. Since there is no report on the link between COPI and callose/PD, it was extremely curious to know the roles of 1-COP or COPI in PD regulation through callose deposition. Here, we report that loss-of-function of 1-COP directly elevates the callose accumulation at PD by affecting subcellular protein localization of callose degradation enzyme PdBG2. This process is linked to ERH1, an inositol phosphoryl ceramide synthase (IPCS), and glucosylceramide synthase (GCS) functions through physical interactions with the 1-COP protein. In addition, the loss-of-function of 1-COP also alters the subcellular localization of ERH1 and GCS proteins, results in a reduction of GlcCers and GlcHCers molecules, which are the key SL species for lipid raft formation. According to our findings, we propose that 1-COP protein, together with the SL modifiers controlling lipid raft compositions, regulates the function of GPI-anchored PD proteins and hence the callose turnover at PD and symplastic movement of biomolecules. Our findings provide the first key clue to link the COPI-mediated intracellular trafficking pathway to the callose-mediated intercellular signaling pathway through PD. One-sentence summaryPlant-specific coatomer protein functions as a negative regulator of callose accumulation by regulating the translocation of sphingolipid enzymes.

cell biology↗

bioRxiv2018

The consensus molecular classification of muscle-invasive bladder cancer

Muscle-Invasive Bladder Cancer (MIBC) is a molecularly diverse disease with heterogeneous clinical outcomes. Several molecular classifications have been proposed, yielding diverse sets of subtypes, which hampers the clinical implications of such knowledge. Here, we report the results of a large international effort to reach a consensus on MIBC molecular subtypes. Using 1750 MIBC transcriptomes and a network-based analysis of six independent MIBC classification systems, we identified a consensus set of six molecular classes: Luminal Papillary (24%), Luminal Non-Specified (8%), Luminal Unstable (15%), Stroma-rich (15%), Basal/Squamous (35%), and Neuroendocrine-like (3%). These consensus classes differ regarding underlying oncogenic mechanisms, infiltration by immune and stromal cells, and histological and clinical characteristics. This consensus system offers a robust framework that will enable testing and validating predictive biomarkers in future clinical trials.

cancer biology↗