Biology subjects

Shimizu, H.

Publications and source records attributed to Shimizu, H..

At least 19 recordsLinked to original sources

bioRxiv2026

STRIPAK associates with Rabex5-Rabaptin5 to promote Rab5-dependent early endosome formation and EGFR transport

His Domain Protein Tyrosine Phosphatase (HD-PTP) is required for multivesicular body (MVB) formation and endosome maturation, with Rab5-GTP levels raised upon HD-PTP depletion (Parkinson et al., 2021. doi: 10/1242/jcs.259192). Mass Spectrometry-based interactomics revealed that HD-PTP depletion increased association of Rabaptin5, a component of the early endosome Rabex5-Rabaptin5 Rab5-GEF, with multiple components of STRIPAK (striatin-interacting phosphatase and kinase). Native IPs and co-expression studies confirmed association between STRIPAK and Rabex5-Rabaptin5, including Strip1-Rabex5 and SLMAP-Rabaptin5 interactions. Consistent with functional Rabaptin5-STRIPAK links, co-depletion of Strip1 or Rabaptin5 with HD-PTP restored normal early endosome positioning and Rab5 distribution. Strip1 depletion impaired formation of central, Rab5-positive early endosomes, reduced cellular Rab5-GTP levels, and prevented the accumulation of internalised EGF-EGFR into centrally located endosomes. SLMAP depletion accentuated the central clustering of early endosomes and other endosomal compartments while also retarding the generation of new, EGFR-containing endosomes and EGFR lysosomal transport. We conclude that STRIPAK is a novel co-assembly factor of Rabex5-Rabaptin5, exerting subunit-specific effects on early endosomes.

cell biology↗

bioRxiv2026

Characterization of central carbon metabolism of Acinetobacter sp. Tol 5 based on 13C-metabolic flux analysis

Acinetobacter sp. strain Tol 5 is an attractive microorganism because of its ability to utilize various aromatic compounds, including toluene. However, because Tol 5 is a non-model microorganism, the utilization of carbon sources in central carbon metabolism remains unclear. In this study, the activity of the central carbon metabolism was characterized using 13C-metabolic flux analysis (13C-MFA). Tol 5 was cultured in synthetic medium using ethanol or acetate as the sole carbon source. In addition, because toluene is degraded into pyruvate and acetyl-CoA via the toluene dioxygenase pathway, this situation was mimicked by culturing with ethanol and lactate as co-substrates. The 13C-MFA confirmed flux through the glyoxylate shunt, which bypasses part of the TCA cycle when either ethanol or acetate was used as the sole carbon source. In ethanol, the flux ratio of isocitrate dehydrogenase at the isocitrate branch point decreased markedly. Unlike acetate, the conversion of ethanol to acetyl-CoA is accompanied by NADH production. Therefore, the observed change in this flux ratio was influenced by the excess NADH generated. Under ethanol-lactate conditions, Tol 5 simultaneously utilized ethanol and lactate. The ethanol consumption rate decreased to less than half of that observed when ethanol was the sole carbon source. Most of the lactate uptake was converted to pyruvate and metabolized via the TCA cycle. The glyoxylate shunt flux was lower than that in ethanol. These metabolic characteristics provide valuable insights into the use of Tol 5 as a next-generation microbial cell factory.

microbiology↗

bioRxiv2026

Axolotl regeneration reveals a dormant cis-regulatory grammar conserved across vertebrate genomes

Unlike most mammals, which lack the capacity to regenerate complex tissues following injury, other vertebrates such as the axolotl rebuild complete limbs throughout life, yet the regulatory mechanisms underlying this striking difference have remained elusive. Here, we define the core cis-regulatory motif grammar driving axolotl limb regeneration and demonstrate that this grammar is conserved within the syntenic neighborhoods of regeneration-gene orthologs in human and mouse genomes, despite being epigenetically sealed in adult mammalian tissues. Integrating this cross-species grammar projection with AlphaGenome, a multimodal genomic AI capable of predicting epigenomic states from long sequence context, we find that the highest-ranking candidate loci are predicted to occupy a state of bivalent dormancy marked by the co-enrichment of poising and repressive histone modifications alongside suppression of transcriptional activity and chromatin accessibility. Systematic in silico motif perturbation further predicts that this dormant state is actively enforced by specific dormancy-stabilizing sequence elements, and that disrupting these elements shifts candidate loci toward a regeneration-competent chromatin configuration. Our findings support a model in which the regenerative blueprint has not been erased from the mammalian genome but locked within it, opening new avenues for understanding the evolution of regenerative competence and for the rational reactivation of latent regenerative programs.

systems biology↗

bioRxiv2026

Integrating virtual cell modeling with patient-derived transcriptomics to uncover target signals in rheumatoid arthritis

Rheumatoid arthritis affects millions of people worldwide, yet a substantial fraction of patients fail to achieve lasting benefit from current therapies. AI-based virtual cell models offer a promising route to simulate the effects of thousands of potential drug targets computationally, but whether their predictions genuinely reflect what happens in the disease-relevant cells of patients undergoing treatment has been unclear. Here, we introduce ImmunoSTATE, a clinically anchored framework that benchmarks empirical and virtual perturbations directly against patient-derived treatment trajectories in pathogenic CXCL13+ synovial T cells from nine patients with rheumatoid arthritis sampled before and after treatment. Empirical genome-wide CRISPR interference screening prioritizes T cell receptor (TCR)-proximal signaling over the measurable components of the JAK-STAT pathway for transcriptomic reversal of the disease signature, a result that remained robust across leave-one-patient-out reanalyses. Comparison of virtual perturbation signals across healthy donor CD4+ T cells and rheumatoid-arthritis patient cells reveals broad preservation of target priorities, together with selected disease-context-enriched shifts, including stronger prioritization of SOX4 in patient cells. ImmunoSTATE provides a clinically anchored framework for evaluating virtual cell models in disease-relevant immune contexts and demonstrates that distinguishing training-set status and cellular context are essential steps in the reliable interpretation of virtual perturbation outputs.

immunology↗

bioRxiv2026

Dose-dependent modeling of combinatorial drug responses stratifies patient survival and reveals therapeutic vulnerabilities in precision oncology

A substantial proportion of cancer patients fail to benefit from their prescribed combination regimens, yet identifying superior alternatives from the vast pharmacological space prior to treatment failure remains an unsolved clinical challenge. Existing computational approaches either rely on multi-omics profiles unavailable in standard oncological practice or reduce drug efficacy to scalar metrics that discard the dose-dependent resolution essential for therapeutic optimization. Here, we present XACT, a hierarchical deep learning framework that reconstructs full dose-dependent drug responses for both monotherapy and drug combinations using only clinically accessible transcriptomic profiles. By leveraging an asymmetric X-Linear Attention mechanism that models second-order interactions between molecular drug substructures and intracellular signaling pathway activities, XACT captures concentration-dependent pharmacodynamics with state-of-the-art accuracy and generalizability to unseen transcriptomic landscapes. When applied to the TCGA pan-cancer cohort, XACT-derived resistance scores were significantly associated with clinical treatment outcomes and stratified overall survival as the strongest independent prognostic factor after multivariate adjustment for tumor stage and cancer type. Systematic virtual screening revealed therapeutic vulnerabilities and nominated alternative regimens for treatment-refractory sarcoma and pancreatic adenocarcinoma. These results establish XACT as a scalable, interpretable, and clinically translatable framework that advances precision oncology from computational prediction toward data-driven therapeutic prescription.

cancer biology↗

bioRxiv2026

Cell-type-resolved Metabolic Flux Inference Reveals Stromal Metabolic Reprogramming Across Human Cardiomyopathies

Metabolic remodeling is a hallmark of cardiomyopathy, yet which cell types bear the metabolic burden and how cell-type-specific contributions are disrupted remain unclear. Here, we developed a cell-type-resolved genome-scale metabolic flux inference pipeline optimized for post-mitotic cardiac tissue by maximizing ATP synthesis rather than biomass production and applied it to a single-nucleus transcriptomic atlas of human cardiomyopathies (78 donors, 869,449 nuclei). Metabolic impairment in dilated cardiomyopathy (DCM) was most profound in stromal cells, whereas myeloid cells exhibited opposing metabolic activation. DCM- associated impairment followed a genotype-dependent severity gradient from structural gene mutations to pathogenic variant-negative (PVneg) cases. PVneg hearts uniquely harbored 24 altered metabolic pathways not significant in any other genotype. These PVneg-specific signatures were independent of clinical severity, indicating a genotype-intrinsic metabolic program. Extending the analysis to arrhythmogenic cardiomyopathy and hypertrophic cardiomyopathy showed that ATP depletion is shared across cardiomyopathy subtypes, whereas metabolic remodeling differed across disease subtypes. Additionally, gene regulatory network analysis linked these alterations to broad transcription factor (TF) dysregulation and pervasive TF-metabolic coupling across all cell types. These findings redefine PVneg DCM as a metabolically distinct entity and reveal conserved stromal metabolic remodeling across cardiomyopathies, providing a framework for genotype-informed mechanistic stratification.

systems biology↗

bioRxiv2026

Pan-Pharmacological Drug-Target Interaction Prediction with 3D-Informed Protein Encoding at Scale

Accurate prediction of drug-target binding affinity across multiple pharmacological endpoints remains challenging, as most deep learning methodologies focus on a single metric and face a trade-off between incorporating structural information and computational throughput. Here we present OmniBind, a multitask framework that resolves both constraints by encoding protein tertiary structures as discrete token sequences and integrating them with amino acid sequence features through a gated fusion mechanism. Trained on over two million compound-protein pairs from BindingDB, OmniBind simultaneously predicts four pharmacological endpoints in a single forward pass, providing a pan-pharmacological profile across multiple affinity endpoints for each compound-target pair. Across adversarial and temporal benchmarks, OmniBind consistently outperforms state-of-the-art architectures, demonstrating that structural encoding captures physicochemical interaction principles rather than ligand-protein co-occurrence patterns. Attention analysis reveals biologically meaningful binding site recognition, with the model selectively weighting the ABL1 gatekeeper residue T315 and responding to its drug-resistance mutation T315I. As a demonstration of practical utility, proteome-wide screening of 20,421 human proteins recovered 85.7% of known clinical targets of clozapine within the top 200 predictions, correctly resolving its pharmacological profile despite structural similarity to clomipramine. OmniBind provides an accurate, structurally informed, and interpretable platform for multi-endpoint drug-target interaction prediction, with applications to lead optimization, off-target safety assessment, and drug repositioning.

bioinformatics↗

bioRxiv2026

OpenMebius2: GUI-based software for 13C-metabolic flux analysis with tracer labeling pattern suggestions for accurate flux predictions

13C-metabolic flux analysis (13C-MFA) is a crucial technique that experimentally determines metabolic flux distribution. Although precision of each flux strongly depends on tracer labeling pattern, its optimization remains challenging. We developed an integrated platform, OpenMebius2, a graphical user interface (GUI)-based software for 13C-MFA that includes a tracer labeling pattern suggestion function to support subsequent experiments. The proposed function leverages metabolic flux distributions and their 95 % confidence intervals obtained using low-cost 13C-labeled substrates to evaluate hypothetical parallel labeling scenarios and predict improvements in flux estimation precision. Availability and implementationThis software runs on Linux, macOS, and Windows. The source code and binary files are available at https://github.com/metabolic-engineering/OpenMebius2 under the PolyForm Noncommercial License 1.0.0.

bioengineering↗

bioRxiv2026

De novo design of therapeutic scFvs and multi-specific engagers from sequence alone

The discovery of functional antibody therapeutics is fundamentally constrained by the astronomical complexity of the sequence landscape and the inefficiency of empirical screening. While structure-based computational design has made significant strides, its reliance on high-resolution co-complex data restricts its utility against the vast undruggable proteome and rapidly evolving pathogens. Here, we introduce IASO, a sequence-centric generative framework for the de novo design of diverse biotherapeutic modalities, including single-chain variable fragments and multi-specific engagers, requiring only antigen sequence as input. IASO integrates an evolutionary-informed generative engine that constructs antigen-steered candidate libraries with a high-fidelity interaction module capturing synergistic physicochemical motifs. We demonstrate the frameworks robustness by identifying novel binders against diverse clinical targets, where IASO successfully discriminates atomic-level mutations to overcome established drug resistance in EGFR and distinguishes single-residue variants of SARS-CoV-2. Furthermore, we extend this approach to the in silico construction of a complex bispecific T-cell engager that exhibits developability profiles comparable to clinically approved therapeutics, including enhanced solubility and minimal immunogenicity. By transforming serendipitous exploration into predictable high-resolution engineering, IASO establishes a scalable foundation for the rapid development of next-generation biopharmaceuticals directly from sequence data alone.

bioinformatics↗

bioRxiv2026

Geometric-Evolutionary Deep Learning Decodes the Human GPCR-Metabolome Interactome and Enables Systematic De-Orphanization

G protein-coupled receptors (GPCRs) are the largest class of drug targets, yet hundreds of orphan GPCRs lack known endogenous ligands, limiting our understanding of human physiology and therapeutic development. Existing computational approaches often fail to generalize to these unseen targets due to a reliance on target-specific priors and linear feature integration. Here, we present G-LEAP, a deep learning framework that learns generalizable principles of GPCR-ligand recognition by synergizing evolutionary protein language models with 3D-aware geometric molecular representations. By implementing a bilinear interaction module, G-LEAP explicitly models non-linear cross-modal interactions and achieves superior generalization on stringent benchmarks, outperforming state-of-the-art methods by 22.4% in error reduction. Crucially, G-LEAP demonstrates robust chemical discrimination by effectively distinguishing active ligands from property-matched physicochemical decoys in the DUDE-Z benchmark and correctly rejecting 67% of hard-negative artifacts prioritized by physics-based docking simulations. Leveraging this capacity, we constructed a comprehensive atlas of over 120 million predicted interactions between 217,000 human metabolites and the GPCR superfamily, which ranked the true endogenous ligand within the top 1% of candidates for 33.7% of known pairs and identified putative orphan ligands validated by significant tissue-specific co-expression with their biosynthetic enzymes. Furthermore, large-scale virtual screening retrieved potent hits with novel chemical scaffolds distinct from known GPCR ligands, demonstrating robust scaffold hopping. G-LEAP thus provides a systematic and biologically validated platform to accelerate de-orphanization and expand the therapeutic chemical space.

systems biology↗

bioRxiv2026

Comparative genomics reveals signatures of distinct metabolic strategies and gene loss associated with Hydra immortality

Hydra is a freshwater cnidarian genus that provides a unique comparative model for aging research, contrasting the immortal H. vulgaris with the aging-inducible H. oligactis. Here, we report a high-quality, chromosome-level genome assembly of H. vulgaris strain AEP.JNIG. Our assembly is comparable in quality to existing resources, facilitating the analysis of genomic diversity across laboratory strains. Epigenomic profiling revealed that gene-body hypermethylation correlates with transcriptional stability and the suppression of spurious transcription in evolutionary conserved genes, suggesting an epigenetic mechanism for genomic integrity. Furthermore, comparative genomics demonstrated that while Hydra conserves fundamental Hallmarks of Aging pathways, the immortal H. vulgaris paradoxically lacks canonical anti-aging genes (e.g., Klotho, NAMPT) found in the aging-inducible H. oligactis. Instead, H. vulgaris exhibits a distinct metabolic signature related to mitochondrial energy production and NTP synthesis. Collectively, our comparative genomics results suggest multiple potential mechanisms associated with the H. vulgaris immortality and the aging traits of H. oligactis, providing novel targets for future functional studies. Significance statementWhy do some organisms age while others appear not to? The freshwater animal Hydra provides a unique opportunity to investigate this question, as closely related species display contrasting aging phenotypes. We generated a high-quality genome assembly for a new strain of a non-aging species and conducted comparative analyses with related strains and an aging species. Even closely related strains can accumulate substantial genetic divergence over time, and stable DNA modification patterns were associated with consistent gene activity, suggesting a mechanism that may help maintain cellular balance. Surprisingly, several well-known longevity genes are present in the aging species but absent in the non-aging one. This suggests that extended lifespan may not simply depend on possessing more "anti-aging" genes, but instead may reflect differences in how core biological processes are organized. Our study provides new insights into the genetic basis of aging and highlights Hydra as a powerful model for understanding longevity.

genomics↗

bioRxiv2026

A Glycaemia-Prolonging Carbohydrate Gel Reduces Hunger and Cognitive Fatigue During Prolonged Esports Play

AbstractProlonged esports play can induce cognitive fatigue by placing sustained demands on executive control, decision-making, and stress regulation. Since the brain relies primarily on glucose as an immediate energy source, we hypothesized that a carbohydrate gel designed to prolong post-ingestion glycaemia stabilizes interstitial glucose, attenuates hunger and cognitive fatigue, and helps preserve performance during extended esports play. Eleven healthy young men completed a randomized, double-blind crossover study with two sessions in which they ingested one of two carbohydrate gels matched for total energy and total carbohydrate but differing in carbohydrate composition (high-fructose corn syrup [HFCS]-dominant vs maltodextrin [MDX]-dominant) before gameplay. Interstitial glucose was monitored continuously, while hunger, salivary cortisol, pupil diameter, executive function, and in-game performance were assessed repeatedly. Compared with the control (HFCS-dominant) condition, the glycaemia-prolonging formulation maintained higher interstitial glucose levels during the later phase of gameplay, blunted the progressive increase in hunger, and attenuated cortisol responses. These physiological and subjective advantages were accompanied by better executive control and more adaptive defensive behaviors in-game, including fewer fouls and more interceptions. In contrast, pupil diameter, an index of global arousal, remained unchanged across conditions, suggesting that benefits were not driven by altered arousal but by improved glycaemic stability and appetite regulation. These findings indicate that a carbohydrate formulation that prolongs glycaemia can reduce hunger-linked cognitive fatigue and support task-relevant behavior during prolonged esports play. This low-burden nutritional strategy may help sustain mental performance during prolonged digital activities that impose high cognitive demands. Highlights- A glycaemia-prolonging gel maintained higher interstitial glucose late in esports play - Hunger and salivary cortisol increased less with the glycaemia-prolonging formulation - Executive control and defensive performance improved (fewer fouls, more interceptions) - Pupil diameter was unchanged, suggesting benefits were not driven by altered arousal - A low-burden, glycaemia-stabilizing strategy may help sustain cognitive stamina

neuroscience↗

bioRxiv2025

Analysis of cancer mutations introduced into the Drosophila Notch Negative Regulatory Region uncovers a diversity of regulatory outcomes

Activating mutations of Notch are drivers of the blood cell cancer, T-ALL, and some solid tumours. The Negative Regulatory Region (NRR) of the extracellular domain (ECD) and the PEST region of the intracellular domain (ICD) are mutation hot spots which can act synergistically in T-ALL. The NRR, comprised of a Heterodimerisation Domain (HD) and three Lin12/Notch repeats (LNR A-C), masks the S2 cleavage site, normally only exposed following ligand binding and cleaved as the first step that ultimately leads to ICD release. Drosophila mutants have played a key role in analysing Notch structure/function but there have been few mutational studies of the NRR. Here we expressed, in S2 cells, over 20 cancer mutations located in the HD, LNR and LNR/HD interface, introduced into Drosophila Notch. Mutations in the HD domain core did not activate, likely due to absence, in Drosophila, of an S1 cleavage within the HD, required for mammalian Notch activity. In contrast, mutations in the LNR/HD interface behaved similarly to T-ALL, activating constitutively with no further ligand-induction and were synergistic with PEST deletion. Mutations of surfaced-exposed residues of LNR-C, also activated constitutively but remained inducible both by ligand and by an intracellular endocytic regulator, Deltex, and were not synergistic with PEST deletions. These mutations caused elevated Notch levels and decreased turnover, suggesting a novel regulatory mechanism. Our results therefore uncover a variety of outcomes arising from perturbations of the NRR and will facilitate the establishment of Drosophila cancer models and the development of mutant-specific approaches to effective therapies.

cell biology↗

bioRxiv2025

Fine modulation of carbon flow in the central carbon metabolism via ribosome-binding site modification in Escherichia coli

Optimization of flux distribution in central carbon metabolism is important to improve the microbial productivity. As the number of precursors required for synthesis differs for each target compound, optimal flux distribution also varies. A library of mutant strains with diverse flux distributions can aid in optimal strain screening. Therefore, in this study, we aimed to construct a library of Escherichia coli strains with stepwise changes in flux distribution by introducing mutations into the ribosome-binding sites of key enzyme genes on its chromosome. We focused on the flux ratios at the glucose-6-phosphate and acetyl-CoA nodes to enhance mevalonate production. Mutations were introduced into the ribosome-binding sites of pgi and gltA to vary the flux ratios of the two pathway branches. Furthermore, a combinatorial repression library comprising 16 strains was constructed by varying pgi and gltA expression at four levels, and a plasmid expressing mevalonate synthesis genes was introduced into each strain. Batch cultures were performed to obtain strains with mevalonate titers and yields were 2.4- and 3.4-fold higher than those of the parent strain. Overall, our combinatorial suppression library of pgi and gltA facilitated the effective identification of mutants with optimal metabolism for mevalonate production.

molecular biology↗

bioRxiv2025

A Functional Atlas of the Tardigrade Resistome Reveals a Diverse Molecular Toolkit for Extremotolerance

How organisms survive environmental extremes that push life to its physical limits is a fundamental question in biology. Tardigrades are a paradigm for this resilience, entering a state of suspended animation called cryptobiosis to withstand near-complete desiccation and intense radiation. While a few key effectors, including the DNA-shielding protein Dsup and various tardigrade-specific intrinsically disordered proteins (TDPs) have been identified, they constitute only a fraction of a vast, uncharacterized proteome, fundamentally limiting a systems-level understanding of this remarkable biology. Here we overcome this limitation by developing AEGIS, an AI-driven engine that systematically discovers and prioritizes novel guardian proteins constituting the tardigrade molecular shield. Applying AEGIS to the tardigrade proteome, we construct the first functional atlas of the tardigrade resistome, revealing hundreds of novel protein families organized into a diverse and compartmentalized molecular toolkit. This atlas pinpoints new Dsup-like nuclear proteins, suggesting multi-layered genome protection, and a rich cohort of cytoplasmic proteins poised to form protective biological glasses. Our work provides a foundational resource for understanding the evolution of extremotolerance and serves as a blueprint for engineering novel biomolecules with transformative biotechnological potential.

bioinformatics↗

bioRxiv2025

A Method for Predicting Enzyme Substrate Specificity Residues Using Homologous Sequence Information

Identifying amino acid residues that are critical for the catalytic function of enzymes is essential for elucidating reaction mechanisms, facilitating drug discovery, and advancing protein engineering. However, experimentally and computationally distinguishing residues that maintain structural integrity from those directly involved in enzymatic function remains a major challenge. In this study, we developed a methodology to identify amino acid residues that influence substrate specificity in enzymes with homologous structures. We framed the sequence comparison as a classification problem, treating each residue as a feature, thereby enabling the rapid and objective identification of key residues responsible for functional differences. To validate the proposed method, we applied it to three enzyme pairs-- trypsin/chymotrypsin, adenylyl cyclase/guanylyl cyclase, and lactate dehydrogenase (LDH)/malate dehydrogenase (MDH). The results confirmed the accurate prediction of previously identified specificity-determining residues. Furthermore, we conducted experiments on the LDH/MDH pair and successfully introduced mutations into key residues to alter substrate specificity, enabling LDH to utilize oxaloacetate while maintaining its expression levels. These findings demonstrate the potential of this method for efficiently identifying residues that govern substrate specificity. We have further developed this approach into a practical tool, the EZSCAN: Enzyme Substrate-specificity and Conservation Analysis Navigator (https://ezscan.pe-tools.com/), which enables rapid identification of amino acid residues critical for enzyme function. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=75 SRC="FIGDIR/small/656053v1_ufig1.gif" ALT="Figure 1"> View larger version (29K): org.highwire.dtl.DTLVardef@d97b76org.highwire.dtl.DTLVardef@38bbe8org.highwire.dtl.DTLVardef@b8b6edorg.highwire.dtl.DTLVardef@f1b858_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2025

High-Resolution Mapping of the Human E3-Substrate Interactome using Ubicon Uncovers Network Architecture and Cancer Vulnerabilities

Mapping the intricate network of E3 ubiquitin ligase-substrate interactions (ESIs), essential for cellular regulation and implicated in numerous diseases including cancer, remains a central challenge limiting mechanistic understanding. Existing experimental and computational methods suffer from limitations in throughput, accuracy, and contextual relevance. Here, we introduce Ubicon, a deep learning framework designed to overcome these hurdles for accurate, proteome-wide human ESI prediction. Ubicon achieves high fidelity by synergistically integrating multimodal features: sequence embeddings from a protein language model specifically adapted for ESI prediction via parameter-efficient fine-tuning (PEFT), predicted 3D structures, and subcellular localization context. Validated extensively, Ubicon achieves state-of-the-art performance (AUROC = 0.9305, AUPRC = 0.6812), significantly surpassing previous approaches. Applying Ubicon, we constructed a high-resolution map of the human E3-substrate interactome, revealing its systems-level architecture characterized by hub proteins and functional modules linked to distinct biological processes. Integrating predictions with cancer genomics further uncovered disease-specific ESI network rewiring, involving oncogenic ligases like AURKA and CDC20, linked to poor prognosis. Ubicon provides a powerful platform to dissect ubiquitin signaling, uncover disease mechanisms, and inform targeted protein degrader development.

bioinformatics↗

bioRxiv2025

MetDeeCINE: Deciphering Metabolic Regulation through Deep Learning and Multi-Omics

Metabolism, the biochemical reaction network within cells, is crucial for life, health, and disease. Recent advances in multi-omics technologies, enabling the simultaneous measurement of transcripts, proteins, and metabolites, provide unprecedented opportunities to comprehensively analyze metabolic regulation. However, effectively integrating these diverse data types to decipher the complex interplay between enzymes and metabolites remains a significant challenge due to the extensive data requirements of kinetic modeling approaches and the limited interpretability of machine learning approaches. Here, we present MetDeeCINE, a novel explainable deep learning framework that predicts the quantitative relationship between each enzyme and metabolite from proteomic and metabolomic data. We demonstrate that our newly developed Metabolism-informed Graph Neural Network (MiGNN), a core component of MetDeeCINE that is guided by the stoichiometric information of metabolic reactions, outperforms other machine learning models in predicting concentration control coefficients (CCCs) using data obtained from kinetic models of E. coli. Notably, MetDeeCINE, even without explicit information on allosteric regulation, can identify key distant enzymes that predominantly control the steady-state concentrations of specific metabolites. Application of MetDeeCINE to mouse liver multi-omics experimental data further demonstrated its ability to generate biologically meaningful predictions through identifying a rate-limiting enzyme of gluconeogenesis associated with obesity, consistent with existing knowledge. MetDeeCINE offers a scalable and interpretable approach for deciphering complex metabolic regulation from multi-omics data, with broad applications in disease research, drug discovery, and metabolic engineering.

systems biology↗