Biology subjects

Moon, S.

Publications and source records attributed to Moon, S..

At least 19 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

SEME enables programmable transcript reconstruction through synthetic microexon incorporation

Precise correction of pathogenic mutations remains challenging for therapeutic applications because genome editing within protein-coding exons can generate unintended insertion-deletion byproducts that disrupt coding integrity and protein function. To address this limitation, we developed the Spliceable Editable Microexon Element (SEME), an intron-targeting platform that enables programmable microexon incorporation through endogenous splicing. In induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs), SEME corrected aberrant splicing caused by the dilated cardiomyopathy-associated FLNC c.2003A>G variant by restoring the five nucleotides missing from exon 12, thereby recovering filamin C expression. Together, these findings establish programmable microexon incorporation as a proof-of-concept strategy for correcting diverse disease-causing transcript defects.

bioengineering↗

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

Motor control of the Drosophila antennae

Animals actively sense their surroundings to acquire behaviorally relevant environmental cues and stimuli. This dynamic acquisition of sensory information is enabled by active positioning of sensors and helps guide behavioral responses in dynamic environments. Yet how these active movements are controlled during behavior and coordinated with ongoing sensory acquisition is not fully understood. In the fruit fly Drosophila melanogaster, the antennae are crucial sensors for extracting important information from the environment including mechanosensory, olfactory, and auditory signals. Just four distinct muscles command movement of the antennae, providing a tractable model system for understanding efferent control of sensation. This work characterizes motor neurons used by Drosophila to actively position the antennae. We first identify antennal motor neurons in the central brain, and map each one from a comprehensive connectomic dataset to its peripheral muscle target. Our analysis of presynaptic inputs to the entire antennal motor system reveals a diverse array of premotor neurons for antennal motor control. We then provide genetic access to each motor unit by building a library of genetic lines with expression in antennal motor neurons. Using this library of antennal motor neuron lines, we next characterize motor unit function with quantitative behavior and optogenetics, revealing that the antennal motor system produces two primary movements in the dorsal-ventral and medial-lateral axes. Together, this work provides a comprehensive framework for understanding the motor control of an active sensor.

neuroscience↗

bioRxiv2026

ASPL-driven subunit exchange remodels VCP/p97 hexamers and is impaired by a multisystem proteinopathy mutation

Valosin-containing protein (VCP/p97) is an essential homohexameric AAA+ ATPase that powers ubiquitin-dependent protein quality control by extraction and unfolding of clients for proteasomal degradation. Heterozygous, autosomal-dominant VCP missense mutants are associated with multisystem proteinopathy (MSP) through unclear molecular mechanisms. We developed a single-molecule pull-down assay to quantify VCP hexamer assembly and subunit exchange dynamics directly in human cell lysate. We show the common MSP-associated VCP variant R155H co-assembles with wild-type subunits to form heterohexamers. Wild-type VCP complexes readily undergo subunit exchange in cell lysates, but this exchange is markedly reduced for purified complexes in buffer. We identify the VCP-interactor ASPL as a selective mediator of monomer exchange that efficiently remodels wild-type, but only modestly exchanges R155H variants within multimers. Single-molecule kinetics analyses reveal [~]2-fold faster ASPL association with, and [~]4-fold slower dissociation from, wild-type VCP than R155H-VCP. We propose that ASPL-driven monomer exchange remodels VCP molecular machines to sustain proteostasis. The failure of ASPL-driven exchange of MSP variant monomers would be predicted to stabilize mutant VCP in assemblies, revealing a potentially targetable defect.

biophysics↗

bioRxiv2026

Npl4 decodes polyubiquitin length and gates D1-D2 coupling in human VCP/p97

VCP/p97 binds the Npl4-Ufd1 heterodimer adaptor to extract polyubiquitinated substrates for proteasomal degradation, but how it decodes K48-linked chain length and how D1-coupled events license downstream D2 power strokes remain unclear. Here we introduce smUbiRAD, or single-molecule ubiquitin recognition and dynamics, and identify a sharp chain-length threshold: Npl4 binds transiently to short chains but switches to long-lived, multivalent engagement on tetra- and penta-ubiquitin. Ufd1 and p97 further stabilize these complexes mainly by suppressing Npl4 dissociation without affecting initial encounter. In fully assembled p97-Ufd1-Npl4-substrate complexes, D1 ATP hydrolysis--rather than D2--drives rapid Npl4 exchange. These results support a model in which D1-powered conformational changes promote cofactor Npl4, but not Ufd1, turnover and gate iterative coupling to downstream D2-driven substrate processing. Finally, we show that multisystem proteinopathy variants R155H and A232E bias p97 toward a high-affinity resting state and accelerate Npl4 exchange, implicating hyperactive cofactor cycling as a disease-linked dysregulation.

biophysics↗

bioRxiv2026

Resolution of recursive data corruption to transform T-cell epitope discovery

Accurate prediction of MHC class I-presented peptides is essential for any vaccine or T-cell therapy design, yet reported gains on in silico benchmarks have not translated into clinical successes. Here we show that this discrepancy may come from a common methodological error: immunopeptidomics datasets are fundamentally contaminated by existing prediction models through prediction-based deconvolution and filtering, resulting in an iterative confirmation bias. An audit of the IEDB, the biggest database in the field, reveals that as of January 2025, 55.8% of assessable data are labeled by computational models rather than verified experimentally. This inflates in silico benchmarks while degrading real-world applicability on new data, effectively making it impossible to objectively test model performance, which can lead to choosing suboptimal solutions and decreasing the chance of any therapys clinical success. In silico simulation shows that iterative data corruption maintains high AUROC while top-of-list retrieval collapses. We reframe epitope discovery as a protein-centric learning-to-rank task and introduce deepMHCflare, a model evaluated exclusively on clean data. deepMHCflare achieves 0.80 Precision@4 on mono-allelic benchmarks versus 0.55-0.65 for gold-standard prediction models. A preclinical cancer vaccine study validated that 2 of the 4 deepMHCflare-nominated peptides were immunogenic, with a third independently confirmed in the literature.

bioinformatics↗

bioRxiv2025

Functional Characterization of 123 Genes from Mycobacteriophage LeBron Uncovers Cytotoxic Gene Families in Cluster L1

Bacteriophages represent a vast reservoir of genetic diversity however, functional annotation remains a major challenge, as most predicted gene products lack detectable similarity to characterized gene families. Experimental approaches such as systematic overexpression screens provide an avenue for identifying phage genes that influence bacterial physiology. Here, we report an overexpression screen of all 123 predicted protein-coding genes from Cluster L1 mycobacteriophage LeBron, the first representative of this cluster to undergo genome-wide functional analysis. Expression assays in Mycobacterium smegmatis revealed that 39 genes (32%) impaired host growth, with nineteen of these toxic genes (49%) assigned no known function. The proportion of cytotoxic genes observed in LeBron is comparable to findings from Clusters K and F, despite minimal sequence conservation across clusters. Interestingly, a subset of LeBrons toxic genes appear to be functionally analogous to previously identified toxic genes in other clusters, suggesting conserved biological activities carried out by non-homologous proteins. Additionally, this analysis uncovered several novel gene families that elicit strong cytotoxic effects, expanding the known catalog of phage-derived bacterial growth inhibitors. These results provide new insights into phage gene functions and demonstrate the value of genome-wide expression screening for uncovering conserved and cluster-specific interactions between bacteriophages and their hosts.

microbiology↗

bioRxiv2025

Network modularity reveals context and state-dependent reorganization of time-varying functional connectivity in single-cell resolved neural activity recordings

An important goal of neuroscience is to understand how biological neural networks organize activity at multiple scales to enable complex information processing and behavioral output. To address this challenge, large-scale neural activity datasets with increased resolution and wider coverage have become more prevalent across many model systems. However, bridging the gap in scale between changes in pairwise functional connectivity between neurons and changes in brain-wide organization of activity remains a key challenge. In this work, we demonstrate application of modularity-based community detection and network modularity to single-cell resolved recordings, for the first time, as a method to summarize complex changes in time-varying functional connectivity, facilitating comparisons across multiple time windows, recordings, and conditions. We apply these methods to both single-cell resolved multi-cell and whole-brain activity recordings. In the multi-cell recordings, we find that food odor changes functional connectivity between existing network modules in a C. elegans locomotory interneuron network, rather than reorganizing them. In spontaneous whole-brain activity, we identify several key hub neurons and combinations that significantly destabilize module assignments when silenced. Together, these results demonstrate community detection and modularity as a method for detecting context and network state-dependent changes in functional connectivity at the intermediate scale of network modules in single-cell resolved neural activity. Results from these analyses facilitate future investigation of mechanisms that mediate organization of neural activity at intermediate scales.

neuroscience↗

bioRxiv2025

Aversive Learning Induces Context-Gated Global Reorganization of Neural Dynamics in Caenorhabditis elegans

Learning generates experience-dependent changes to the brain. However, how neurons of diverse functions and connectivity reorganize and modulate their activities to generate coherent changes while preserving essential functions is not well understood. Here, we address this question using an aversive olfactory learning paradigm whereby Caenorhabditis elegans learns to reduce its olfactory preference for pathogenic bacteria. Using functional imaging during olfactory stimulation in naive and trained animals, we show that, at brain-wide scale, cell type-by-cell type, learning induces context-gated reconfiguration of the organization of neural activity throughout the brain to alter neural responses only during bacteria-discrimination task, while leaving intact bacteria-sensing functions. We found that the context-gated encoding of learning is globally distributed across layers of the nervous system, composed of neurons carrying information of context or learning experience. In particular, aversive training modulates multiple functional connections within the nervous system, including those between sensory neurons and interneurons, as well as those among interneurons, in a context-gated manner. At the systems level, training modulates correlated activity of neural populations; we found that low-dimensional temporal patterns of population activity correlate well with locomotory gaits that express olfactory preferences. Upon training, the rotation and contraction of the low-dimensional neural manifolds shifts the brain into predisposed states for the context-gated display of learned behavior. Because animals encounter unpredictable environments in life, efficient learning about relevant cues while maintaining other functions is essential for survival. Our findings uncover network-level mechanisms that help explain how the brain reorganizes its activity patterns to both encode new experience and preserve essential functions.

neuroscience↗

bioRxiv2025

Toward Single-Cell Control: Noise-Robust Perfect Adaptation in Biomolecular Systems

Robust perfect adaptation (RPA), whereby a consistent output level is maintained even after a disturbance, is a highly desired feature in biological systems. This property can be achieved at the population average level by combining the well-known antithetic integral feedback (AIF) loop into the target network. However, the AIF controller amplifies the noise of the output level, disrupting the single-cell level regulation of the system output and compromising the conceptual goal of stable output level control. To address this, we introduce a new regulation motif, the noise controller, which is inspired by the AIF loop but differs by sensing the output levels through the dimerization of output species. Combining this noise controller with the AIF controller successfully maintained system output noise as well as mean at their original level, even after the perturbation, thereby achieving noise RPA. Furthermore, our noise controller could reduce the output noise to a desired target value, achieving a Fano factor as small as 1, the commonly recognized lower bound of intrinsic noise in biological systems. Notably, our controller remains effective as long as the combined system is ergodic, making it applicable to a broad range of networks. We demonstrate its utility by combining the noise controller with the DNA repair system of Escherichia coli, which reduced the proportion of cells failing to initiate the DNA damage response. These findings enhance the precision of existing biological controllers, marking a key step toward achieving single-cell level regulation.

systems biology↗

bioRxiv2025

High speed functional imaging with a microfluidics-compatible open-top light-sheet microscope enabled by model predictive control of a tunable lens

1Functional fluorescence imaging of small, transparent organisms has proven to be a powerful tool for understanding the development and function of intact nervous systems at the cellular level, particularly when combined with microfluidic tools that enable precise manipulation of sensory cues. Light-sheet fluorescence microscopy has a number of advantages for functional imaging, including efficient axial sectioning, reduced photobleaching, and high speed compared to confocal methods, but many configurations with ideal properties for efficient functional imaging place constraints on sample access and geometry that preclude their use with microfluidics. We present an open-top light-sheet microscope that uses a simplified inverted water immersion interface coupled with tunable lens remote focusing to achieve high-speed, multichannel 3D imaging of specimens in conventional microfluidics. We use model predictive control to efficiently optimize drive signals for an electrically tunable lens, providing reliable, camera-limited scanning of the imaging volume. We then demonstrate the utility of this approach by recording calcium activity from C. elegans at volume rates up to 20 Hz, including both whole-brain response to chemical stimulation and compartmentalized dendritic response of the PVD neuron to mechanical stimulation. Our approaches are flexible and inexpensive to implement, and could be adapted to improve performance and sample compatibility for a wide range of imaging techniques.

neuroscience↗

bioRxiv2025

A Precision Ultrasound-Localized Sonoporation-Equipment (PULSE) Enabling Intratumoral Delivery for Cancer Immunotherapy

Cancer immunotherapy offers a promising long-term treatment for solid tumors by activating the immune system, but side effects and systemic toxicity limit its broad clinical translation. While intratumoral delivery addresses these limitations, elevated stiffness within solid tumors continues to hinder both intracellular and extracellular delivery. Here, a novel therapeutic platform, the Precision Ultrasound-Localized Sonoporation-Equipment (PULSE), is presented to enable spatiotemporally controlled, ultrasound-mediated gene and drug delivery for intratumoral immunotherapy. The PULSE incorporates forward and sideward-looking miniaturized ultrasound transducers, integrated into either a 7-French catheter (cPULSE) or a 16-gauge needle (nPULSE), to minimally invasively sonicate a centimeter-sized tumor. Following the design, prototyping, and characterization of PULSE, its dual capability for intracellular and extracellular delivery is demonstrated through comprehensive in vitro cell-based sonoporation tests and phantom-based drug penetration tests. Luciferase assays (activity: ~104 RLU/g level) in the cell studies confirm significantly enhanced gene transfection with cPULSE, while increased dye diffusion (width: ~1 cm level) in the phantom tests validates improved perfusion with nPULSE. The reported PULSE shows promise for spatiotemporally precise, controlled, and localized therapeutic delivery in early-stage tumors for intratumoral immunotherapy.

bioengineering↗

bioRxiv2025

Use of a specific set of learner-centered evidence-based teaching practices correlates with higher exam performance across seven STEM departments

Myriad studies support the claim that active learning improves student academic performance in STEM, yet lecture remains the dominant form of instruction. Faculty offer multiple reasons for not using active learning with many expressing confusion as to what active learning is. Contributing to that confusion is the fact there are a multitude of named active learning methods and varied implementations of each. In an effort to better understand how specific elements of active learning might contribute to enhanced academic performance, we used a more fine-grained classroom observation tool, PORTAAL, to observe teaching practices across 157 STEM courses. We used a principal component analysis to reduce the observed teaching practices to specific sets of practices. We found a continuum of implementation of teaching practices that ranged from instruction-centered to learner-centered. The instruction-centered practices included high Blooms level questions, students working alone, instructors answering and explaining questions, and providing alternative answers. The learner-centered practices were volunteer or randomly-called students explaining answers, instructors prompting students logic, and instructors giving positive feedback. Using a linear regression to analyze the data from all courses, we found medium and medium-high levels of learner-centered teaching practices correlated with higher student exam performance compared to instruction-centered practices. We observed a similar pattern when analyzing data only from introductory courses. We also analyzed interactions of both binary gender and first generation status with learner-centered teaching on exam scores. We propose that when instructors use the learner-centered practices, they shift the responsibility of the intellectual work of the problem-the learning-to the students, which may result in improved exam performance. Even moderate levels of these learner-centered practices have a positive correlation with exam performance. Faculty can incorporate these key elements of learner-centered practices into their own teaching based on their comfort level, course content, and abilities of their students.

scientific communication and education↗

bioRxiv2025

tRNA synthetase activity is required for stress granule and P-body assembly

In response to stress, translation initiation is suppressed and ribosome runoff via translation elongation drives mRNA assembly into ribonucleoprotein (RNP) granules including stress granules and P-bodies. Defects in translation elongation activate the integrated stress response. If and how stalled ribosomes are removed from mRNAs during translation elongation stress to drive RNP granule assembly is not clear. We demonstrate the integrated stress response is induced upon tRNA synthetase inhibition in part via ribosome collision sensing. However, saturating levels of tRNA synthetase inhibitors do not induce stress granules or P-bodies and prevent RNP granule assembly upon exogenous stress. The loss of tRNA synthetase activity causes persistent ribosome stalls that can be released with puromycin but are not rescued by ribosome-associated quality control pathways. Therefore, tRNA synthetase activity is required for ribosomes to run off mRNAs during stress to scaffold cytoplasmic RNP granules. Our findings suggest ribosome stalls can persist in human cells and uniquely uncouple ribonucleoprotein condensate assembly from the integrated stress response.

molecular biology↗

bioRxiv2024

Flotillin-2 dampens T cell antigen-sensitivity and functionality

T cell receptor (TCR) engagement triggers T cell responses, yet how TCR-mediated activation is regulated at the plasma membrane remains unclear. Here, we report that deleting the membrane scaffolding protein Flotillin-2 (Flot2) increases T cell antigen sensitivity, resulting in enhanced TCR signaling and effector function to weak TCR stimulation. T cell-specific Flot2-deficient mice exhibited reduced tumor growth and enhanced immunity to infection. Flot2-null CD4+ T cells exhibited increased T helper 1 polarization, proliferation, Nur77 induction, and phosphorylation of ZAP70 and LCK upon weak TCR stimulation, indicating a sensitized TCR-triggering threshold. Single cell-RNA sequencing suggested that Flot2-null CD4+ T cells follow a similar route of activation as wild-type CD4+ T cells but exhibit higher occupancy of a discrete activation state under weak TCR stimulation. Given prior reports that TCR clustering influences sensitivity of T cells to stimuli, we evaluated TCR distribution with super-resolution microscopy. Flot2 ablation increased the number of surface TCR nanoclusters on naive CD4+ T cells. Collectively, we posit that Flot2 modulates T cell functionality to weak TCR stimulation, at least in part, by regulating surface TCR clustering. Our findings have implications for improving T cell reactivity in diseases with poor antigenicity, such as cancer and chronic infections.

immunology↗

bioRxiv2023

Allelic Imbalance Reveals Genetic and Epigenic Underpinnings of metabolic diseases

BackgroundObesity is a pivotal trigger for a spectrum of complex metabolic disorders. By colocalizing cis-eQTLs in adipose tissues from the GTEx consortium and trait-associated SNPs for complex traits from the GWAS Catalog within 3.6 million DNase I hypersensitive sites (DHSs), we systematically identify regulatory variants and genes that exhibit cis effects, as well as potential causal variants within the context of regulatory elements. ResultsOur analysis reveals that 229,504 (26.4%) cis-eQTLs located within DHS reside densely near the transcription start sites, contrasting with those outside of DHS. We observed that genes with higher allelic imbalance have shorter transcript lengths with larger number cis-eQTLs within DHS, and such imbalance genes are predominantly linked to signaling and immune response, whereas those with lower allelic imbalance tend to be involved in metabolism. Our composite colocalization score prioritizes 5,202 DHSs that encompass both cis-eQTLs and trait-associated SNPs, targeting 2,232 protein-coding genes and 523 lncRNAs across complex traits. We highlight the lncRNA SNHG5 as a prime example; it is associated with high-density lipoprotein levels and exhibits low allelic imbalance, and is also down-regulated in adipose tissue from patients with obesity. ConclusionsOur findings underscore the critical role of regulatory context in pinpointing causal variants and refining target genes, offering rich insights into the genetic mechanisms pertinent to obesity and providing valuable resources for the diagnosis and therapeutic targeting of metabolic diseases.

genomics↗