Biology subjects

Mu, Y.

Publications and source records attributed to Mu, Y..

At least 19 recordsLinked to original sources

bioRxiv2026

Filling the Missing Puzzle: Integrating Interaction and Evolutionary Information to Enhance Protein Fitness Prediction in Inverse Folding

Protein inverse folding, which infers compatible amino acid sequences from a given backbone structure, is central to rational protein design. Current methods rely largely on sequence recovery rate as the evaluation metric, emphasizing structural compatibility but seldom addressing function. Yet functional activity depends on the proper network of non-covalent interactions-protein-ligand contacts being a prime example. We argue that function-oriented protein design is governed by three complementary constraints: structural, interaction, and evolutionary, none of which are handled in a unified manner by existing pretrained models. To bridge this gap, we introduce zPolaron, a graph neural network built on protein-ligand complexes for functional protein design. zPolaron represents residues and ligand atoms as a heterogeneous graph and jointly encodes structural, interaction, and evolutionary information by fusing physical-interaction-enhanced graph representations with ESM-2 embeddings. Across multiple fitness prediction benchmarks, zPolaron outperforms all baselines, with gains evident across structural, interaction, and evolutionary dimensions. Wet-lab experiments further confirm its utility in functional enzyme screening and design: the optimized CalA6-D205A-S160G mutant achieves a 17.84-fold improvement in activity, while the CalB-I121M mutant yields a 1.91-fold increase. Together, these results show that integrating interaction and evolutionary information fills the missing puzzle of inverse folding, enhancing protein fitness prediction beyond structural compatibility. zPolaron is available at https://github.com/zelixirSH/zPolaron.

bioinformatics↗

bioRxiv2026

Decoding the Transcriptome Dark Matter: Construction of Single-Cell Whole-Transcriptome Regulatory Atlas by dropTotal

To profile unknown ncRNAs-"dark matter" in single cells, we developed dropTotal, a high-throughput droplet-based total RNA-seq method that uses dU-modified GAT primer with temperature-ramp hybridization and droplet merge barcoding to co-detect coding and non-coding transcripts with record sensitivity (>13,500 genes/cell, including >2,000 lncRNAs and >500 sncRNAs), compatible with fresh, frozen, fixed, and FFPE tissues. Applied to ~75,000 human glioma nuclei, it captured 60,313 genes (18,681 lncRNA, 19,859 mRNAs and 6,753 sncRNAs), enabling ncRNA-driven regulatory landscape construction. In oligodendroglioma, module analysis identified recurrence-associated ncRNA-centered modules linked to therapy resistance and invasion; in glioblastoma, six cellular states showed hundreds of state-specific unannotated ncRNAs with divergent functions, from MIR222HG-mediated immune modulation to SCIRT-driven hypoxia adaptation. Alternative splicing analysis identified 428 state-specific junction markers and mapped cell-state-specific alternative splicing regulation. dropTotal offers broad application for decoding the underlying ncRNA biology and single-cell whole transcriptome regulatory mechanisms in cellular identity and disease progression.

genomics↗

bioRxiv2026

CrossAffinity: A Sequence-Based Protein-Protein Binding Affinity Prediction Tool Using Cross-Attention Mechanism

Protein-protein binding affinity is important for understanding protein interactions within a protein complex and for identifying strong drug-peptide binders to a target protein. Many structure-based models were built previously with reasonable performance. However, such models require protein complex structure as input, which is usually unavailable due to high cost and experimental constraints. To tackle such an issue, the sequence-based CrossAffinity model was constructed in this study, using the cross-attention module to extract contextual information of interacting protein components while separating the protein complex into two distinct parts to predict the protein-protein binding affinity. CrossAffinity managed to outperform all structure-based models and sequence-based models in an S34 test set containing newer protein complex structures and binding affinity values in a timeline while being trained on an older dataset, showing generalisability to new data points. In other test sets, namely S90, S90 subset and S79*, CrossAffinity also managed to outperform all other sequence-based models while maintaining comparable performance to many recently published structure-based models. The acceptable performance and quick inference of CrossAffinity enable it to be deployed in situations requiring the prediction of the binding affinity of many protein complexes that lack structural information.

bioinformatics↗

bioRxiv2026

Metadiffusion: inference-time meta-energy biasing of biomolecular diffusion models

Biomolecular function often depends on conformational ensembles, yet modern diffusion-based structure generators are biased toward the compact conformations prevalent in structural databases, limiting their ability to explore broad conformational landscapes. This work introduces metadiffusion, where an additional meta-energy biasing layer on top of diffusion steers pretrained biomolecular diffusion models through gradient-guided denoising. Without retraining, metadiffusion generates diverse conformational ensembles whose residue-level flexibility patterns closely match molecular dynamics simulations. The method supports three complementary modes: optimisation, steering to user-specified targets, and exploration via inter-sample repulsion. This approach enables controlled exploration of collective variables, enumeration of alternative binding poses across proteins, nucleic acids and ligands, and conformational ensemble generation consistent with SAXS and NMR chemical shifts. Metadiffusion thus provides a practical route to connect diffusion-based structure generation with ensemble-level, experimentally-restrained structural analysis.

bioinformatics↗

bioRxiv2025

Neural ensembles for music production recruit more language instruments as rhythmic complexity increases

Language and music, two core forms of human communication, both rely on rhythmic control. To date, the neural ensembles supporting their real-time production are not fully characterized. Using rapid phase-encoded fMRI, we mapped the spatiotemporal brain dynamics of overt language and music production under varying rhythm control demands. We observed overlapping hemodynamic traveling waves across domains, with auditory-motor regions activated earlier, followed by visual, posterior parietal, and Sylvian parietal-temporal (Spt) regions, supporting sensorimotor transformation and online monitoring of production. Increasing rhythmic complexity elicited slower and stronger activations in both domains: compared with baseline, regular rhythm control elicited delayed and amplified responses, while irregular rhythm control produced the slowest and strongest responses. During music production, rhythm-related activity expanded into frontal and temporal language regions, suggesting that higher rhythmic demands engage language resources. These findings reveal dynamic neural ensembles that can flexibly reconfigure shared resources to support temporal precision in sequence production.

neuroscience↗

bioRxiv2025

Cohesin Acetylation and ATPase Activity Control Cohesion and Loop Architecture through Distinct Mechanisms

Cohesin is a conserved protein complex that mediates sister chromatid cohesion, chromosome condensation, gene regulation, and DNA repair. These processes rely on cohesins ability to tether sister chromatids and form chromatin loops, which depend on cohesins ATPase activity and Eco1-mediated acetylation of two lysine residues (K112 and K113 in budding yeast) in its Smc3 subunit. How cohesins ATPase activity and acetylation integrate to control cohesin functions remains poorly understood. Here, we analyzed chromatin architecture in yeast mutants with altered cohesin acetylation and/or ATPase activity. We find that acetylation of either K112 or K113 is sufficient to produce a wild-type chromosome structure with loops positioned at cohesin-associated regions (CARs), whereas loss of acetylation at both residues abolishes positioned loops, indicating that acetylation at either lysine alone can maintain wild-type chromatin architecture. We further show that a cohesin acetylation mutant, despite being defective in sister-chromatid tethering and thus failing to establish cohesion, still forms wild-type-like loops, while cohesion-competent mutants lack positioned loops. These results suggest that the activities required for cohesion and loop formation are mechanistically separable, arguing against passive loop capture. Moreover, a mutant with reduced ATPase activity showed a loop profile similar to wild type, indicating that cohesin with lower ATPase activity can still form wild-type chromatin architecture. By contrast, hyper-ATPase mutants accumulate positioned loops, suggesting that increasing ATPase activity can enhance loop processivity. Together, our findings support a multilayered regulatory model in which acetylation fine-tunes ATPase output and cohesin functions to shape genome architecture.

genetics↗

bioRxiv2025

Reproducibility of Scientific Claims in Drosophila Immunity: A Retrospective Analysis of 400 Publications

Drosophila immunity has been the focus of intense study and has impacted other research fields including innate immunity and agriculturally or epidemiologically relevant investigations of insect pests and vectors. Unsurprisingly for such a large body of work, some published results were later found to be irreplicable. Although some results have been contradicted in the literature, many have no published follow-up, either due to a lack of research or low motivation to publish negative or contradictory results. We have addressed this by performing a reproducibility project that analyses the conceptual replicability of claims from articles published on Drosophila immunity before 2011. To assess replicability, we extracted claims from 400 articles on the Drosophila immune response to bacteria and fungi and performed preliminary verification by comparing these claims to other published literature in the field. Using alternative approaches, we also experimentally tested some unchallenged claims, with no published follow-up. The intent of this analysis was to centralize evidence on insights and findings to improve clarity for scientists that may base research programs on these data. Although the aim of the ReproSci project is to assess the replicability of claims made in articles published in the field of Drosophila immunity, it is in no way an assessment of the scientific value of the research. All our data are published on a publicly available website associated with this article (https://ReproSci.epfl.ch/) that encourages community participation. This article provides a short summary of claims that were found to have contradictory evidence, which may help the community to assess past findings on Drosophila immunity and improve clarity going forward. Statistical analysis of this reproducibility study and metascience insights obtained from this approach are discussed in a companion article.

genetics↗

bioRxiv2025

Antigen-specific Th17 T cells offset the age-related decline in durable T cell immunity

Older adults are susceptible to infections, in part due to waning of immune memory. To determine mechanisms that determine long-lasting versus short-term immunity, we examined varicella zoster virus (VZV) vaccination as a model system. We contrasted VZV antigen-specific T cells several years after vaccination in adults who had been vaccinated at young (<20 years) or older age (>50 years) with a live-attenuated vaccine that confers durable protection only when given at young age, or with an adjuvanted VZV component vaccine that elicits effective, long-lasting immunity in older adults. CD8+ T cells were highly sensitive to age-related changes showing T cell subset shifts, loss in TCR diversity and reduced stem-like features while gaining NK-like signatures without evidence for cellular senescence or exhaustion. VZV-specific CD4+ T cells were largely resilient to age and maintained phenotypic and TCR diversity. Immunization of older adults with the adjuvanted VZV vaccine did not reverse age-associated defects in CD8+ T cells. Instead, it selectively improved the functionality of VZV-specific Th17 CD4+ T cells and prevented their acquisition of Treg features, likely as consequence of lipid metabolic pathways. Collectively, our data indicate that effective vaccination in older adults is supported by the generation of a durable, antigen-specific CD4+ Th17 population that resists mis-differentiation into Tregs and that compensates for age-related defects in CD8+ T cells. One sentence summaryAging primarily impairs the CD8+ T cell VZV vaccine response, while effective VZV vaccination in older adults induces antigen-specific Th17 CD4+ T cells to compensate for aging defects.

immunology↗

bioRxiv2025

Central nervous system atlas of larval zebrafish constructed using the morphology of single excitatory and inhibitory neurons

Single-neuron morphology mapping is fundamental for deciphering brain architecture and function, yet in vertebrates it remains limited in spatial coverage and cell-type information. Here we present a nervous system-wide single-neuron morphology atlas of larval zebrafish with annotated excitatory/inhibitory identity and dendrite-axon polarity. The atlas comprises >20,000 reconstructed neurons from >13,400 animals, spanning the central and peripheral nervous system and representing [~]24.5% of the total excitatory and inhibitory neurons in the brain, all integrated into a common physical space with neuroanatomical parcellations. We define >500 excitatory and inhibitory neuronal morphotypes and construct putative inter-cell-type and interregional connectomes. Projection and morphotype analyses reveal distinct laterality and directionality biases and modular innervation patterns in excitatory and inhibitory neurons, while network analyses uncover fundamental connectivity motifs, hubs, and sensorimotor pathways. Accessible through custom-built interactive platform ZExplorer, this atlas provides a foundational resource for multimodal data integration, circuit analysis, computational modeling, and hypothesis-driven research.

neuroscience↗

bioRxiv2025

Harmonizing Heterogeneous Single-Cell Gene Expression Data with Individual-Level Covariate Information

MotivationThe growing availability of single-cell RNA sequencing (scRNA-seq) data highlights the necessity for robust integration methods to uncover both shared and unique cellular features across samples. These datasets often exhibit technical variations and biological differences, complicating integrative analyses. While numerous integration methods have been proposed, many fail to account for individual-level covariates or are limited to discrete variables. ResultsTo address these limitations, we propose scINSIGHT2, a generalized linear latent variable model that accommodates both continuous covariates, such as age, and discrete factors, such as disease conditions. Through both simulation studies and real-data applications, we demonstrate that scINSIGHT2 accurately harmonizes scRNA-seq datasets, whether from single or multiple sources. These results highlight scINSIGHT2s utility in capturing meaningful biological insights from scRNA-seq data while accounting for individual-level variation. Availability and implementationThe scINSIGHT2 method has been implemented as an R package, which is available at https://github.com/yudimu/scINSIGHT2/.

bioinformatics↗

bioRxiv2025

A Sensitive Soma-localized Red Fluorescent Calcium Indicator for Multi-Modality Imaging of Neuronal Populations In Vivo

Recent advancements in genetically encoded calcium indicators, particularly those based on green fluorescent proteins, have optimized their performance for monitoring neuronal activities in a variety of model organisms. However, progress in developing red-shifted GECIs, despite their advantages over green indicators, has been slower, resulting in fewer options for end-users. In this study, we explored topological inversion and soma-targeting strategies, which are complementary to conventional mutagenesis, to re-engineer a red genetically encoded calcium indicator, FRCaMP, for enhanced in vivo performance. The resulting sensors, FRCaMPi and soma-targeted FRCaMPi (SomaFRCaMPi), exhibit up to 2-fold higher dynamic range and peak {Delta}F/F0 per single AP compared to widely used jRGECO1a in neurons in culture and in vivo. Compared to jRGECO1a and FRCaMPi, SomaFRCaMPi reduces erroneous correlation of neuronal activity in the brains of mice and zebrafish by two- to four-fold due to diminished neuropil contamination without compromising the signal-to-noise ratio. Under wide-field imaging in primary somatosensory and visual cortex in mice with high labeling density (80-90%), SomaFRCaMPi exhibits up to 40% higher SNR and decreased artifactual correlation across neurons. Altogether, SomaFRCaMPi improves the accuracy and scale of neuronal activity imaging at single-neuron resolution in densely labeled brain tissues due to a 2-3-fold enhanced automated neuronal segmentation, 50% higher fraction of responsive cells, up to 2-fold higher SNR compared to jRGECO1a. Our findings highlight the potential of SomaFRCaMPi, comparable to the most sensitive soma-targeted GCaMP, for precise spatial recording of neuronal populations using popular imaging modalities in model organisms such as zebrafish and mice.

neuroscience↗

bioRxiv2024

In vivo multiplex imaging of dynamic neurochemical networks with designed far-red dopamine sensors

Neurochemical signals like dopamine (DA) play a crucial role in a variety of brain functions through intricate interactions with other neuromodulators and intracellular signaling pathways. However, studying these complex networks has been hindered by the challenge of detecting multiple neurochemicals in vivo simultaneously. To overcome this limitation, we developed a single-protein chemigenetic DA sensor, HaloDA1.0, which combines a cpHaloTag-chemical dye approach with the G protein-coupled receptor activation-based (GRAB) strategy, providing high sensitivity for DA, sub-second response kinetics, and an extensive spectral range from far-red to near-infrared. When used together with existing green and red fluorescent neuromodulator sensors, Ca2+ indicators, cAMP sensors, and optogenetic tools, HaloDA1.0 provides high versatility for multiplex imaging in cultured neurons, brain slices, and behaving animals, facilitating in-depth studies of dynamic neurochemical networks.

neuroscience↗

bioRxiv2024

A Palette of Bridged Bicycle-Strengthened Fluorophores

Organic fluorophores are the keystone of advanced biological imaging. The vast chemical space of fluorophores has been extensively explored in seek of molecules with ideal properties. However, within the current molecular constraints, there appears to be a trade-off between high brightness, robust photostability, and tunable biochemical properties. Herein we report a general strategy to systematically boost the performance of donor-acceptor-type fluorophores by leveraging SO2 and O-substituted azabicyclo[3.2.1] octane auxochromes. These bicyclic heterocycles give rise to a collection of Bridged dyes (BD) spanning the UV and visible range with top-notch quantum efficiencies, enhanced water solubility, and tunable cell-permeability. Notably, these azabicyclic fluorophores showed remarkable photostability than its tetramethyl or azatidine analogue, at the same time completely resistant to oxidative photobluing rendered by the Bredts rule. Functionalized BD dyes are tailored for applications in single-molecule imaging, super-resolution imaging (STED and SIM) in fixed or live mammalian cells and plant cells, and live zebrafish imaging or chemigenetic voltage imaging. Synergizing with advanced imaging methods, the bridge bicycle dyes represent a versatile palette for biological researches.

biophysics↗

bioRxiv2024

Pharmacophore-constrained de novo drug design with diffusion bridge

De novo design of bioactive drug molecules with potential to treat desired biological targets is a profound task in the drug discovery process. Existing approaches tend to leverage the pocket structure of the target protein to condition the molecule generation. However, even the pocket area of the target protein may contain redundant information since not all atoms in the pocket is responsible for the interaction with the ligand. In this work, we propose PharmacoBridge, a phamacophore-guided de novo design approach to generate drug candidates inducing desired bioactivity via diffusion bridge. Our method adapts the diffusion bridge to effectively convert pharmacophore arrangements in the spatial space into molecular structures under the manner of SE(3) -equivariant transformation, providing sophisticated control over optimal biochemical feature arrangements on the generated molecules. PharmacoBridge is demonstrated to generate hit candidates that exhibit high binding affinity with potential protein targets.

bioinformatics↗

bioRxiv2024

Chemosensory modulation of eye-body coordination in larval zebrafish

Coordinated eye-body movements are essential for adaptive behavior, yet little is known about how multisensory input, particularly chemosensory cues, shapes this coordination. Using our enhanced Fish-on-Chips optofluidic platform, we uncovered complex dynamics in how larval zebrafish coordinate saccadic eye movements with tail flips. Under baseline conditions, spontaneous tail flips dynamically align with saccades in frequency and direction for coordinated turns. Chemosensory valence further modulates this coordination: death-associated cues intensify both the strength and relative frequency of coupled saccade-tail flips during turns, whereas food-related cues promote forward gliding without altering saccade coupling. Concurrent brain-wide neuronal imaging reveals that the zebrafish pallium represents the mapping between aversive valence and enhanced saccade-tail coordination, with stronger coupling associated with higher pallium activation. These findings uncover the neural representations and behavioral strategies by which chemosensory inputs of different valences are associated with amplified eye-body coordination and adaptive locomotion in a developing vertebrate, offering insights into principles of sensory-motor integration.

neuroscience↗

bioRxiv2024

UNG-RPA interaction governs the choice between high-fidelity and mutagenic uracil repair

Mammalian Uracil DNA glycosylase (UNG) removes uracils and initiates high-fidelity base excision repair to maintain genomic stability. During B cell development, activation-induced cytidine deaminase (AID) creates uracils that UNG processes in an error-prone fashion to accomplish immunoglobulin (Ig) somatic hypermutation (SHM) or class switch recombination (CSR). The mechanism that governs high-fidelity versus mutagenic uracil repair is not understood. The B cell tropic gammaherpesvirus (GHV) encodes a functional homolog of UNG that can process AID induced genomic uracils. GHVUNG does not support hypermutation, suggesting intrinsic properties of UNG influence repair outcome. Noting the structural divergence between the UNGs, we define the RPA interacting motif as the determinant of mutation outcome. UNG or RPA mutants unable to interact with each other, only support high-fidelity repair. In B cells, transversions at the Ig variable region are abated while CSR is supported. Thus UNG-RPA governs the generation of mutations and has implications for locus specific mutagenesis in B cells and deamination associated mutational signatures in cancer.

immunology↗

bioRxiv2024

Protein language models are performant in structure-free virtual screening

1.1Hitherto virtual screening has been typically performed using a structure-based drug design paradigm. Such methods typically require the use of molecular docking on high-resolution three-dimensional structures of a target protein - a computationally-intensive and time-consuming exercise. This work demonstrates that by employing protein language models and molecular graphs as inputs to a novel graph-to-transformer cross-attention mechanism, a screening power comparable to state-of-the-art structure-based models can be achieved. The implications thereof include highly expedited virtual screening due to the greatly reduced compute required to run this model, and the ability to perform early stages of computer-aided drug design in the complete absence of 3D protein structure.

bioinformatics↗

bioRxiv2024

Mechanical forces pattern endocardial Notch activation via mTORC2-PKC pathway

Notch signaling has been identified as a key regulatory pathway in patterning the endocardium through activation of endothelial-to-mesenchymal transition (EMT) in the atrioventricular canal (AVC) and proximal outflow tract (OFT) region. However, the precise mechanism underlying Notch activation remains elusive. By transiently blocking the heartbeat of E9.5 mouse embryos, we found that Notch activation in the arterial endothelium was dependent on its ligand Dll4, whereas the reduced expression of Dll4 in the endocardium led to a ligand-depleted field, enabling Notch to be specifically activated in AVC and OFT by regional increased shear stress. The strong shear stress altered the membrane lipid microdomain structure of endocardial cells, which activated mTORC2 and PKC and promoted Notch1 cleavage even in the absence of strong ligand stimulation. These findings highlight the role of mechanical forces as a primary cue for endocardial patterning and provide insights into the mechanisms underlying congenital heart diseases of endocardial origin.

developmental biology↗