Biology subjects

Liu, T.

Publications and source records attributed to Liu, T..

At least 37 records · Page 2Linked to original sources

bioRxiv2026

A functional genomics screen of human B-cell differentiation reveals convergent mechanisms of inherited childhood leukemia predisposition

B-cell acute lymphoblastic leukemia (B-ALL) is the most common childhood cancer, yet the mechanisms by which inherited risk variants predispose to leukemia development remain poorly understood. A major challenge to studying these mechanisms has been the lack of model systems that faithfully capture the transient developmental states in which predisposition alleles are thought to act. Here, we establish a human B-cell differentiation platform from hematopoietic stem/progenitor cells that enables CRISPR-based engineering, recapitulates early B-cell lymphopoiesis, and enriches for rare developmental intermediates. By applying systematic perturbations with multiplexed single-cell transcriptomic profiling to mimic the effects of mutations in nine familial B-ALL predisposition genes, we decipher mechanisms by which B-cell development can be altered by such inherited variation to predispose to B-ALL. Through these studies, we identify convergent delays in B-cell differentiation at progenitor stages characterized by high-level RAG1/2 recombination activity. We propose that these delays at progenitor stages increase the likelihood that cells can undergo illegitimate RAG-mediated recombination to promote transformation, a finding consistent with similar rates of illegitimate RAG-associated genomic alterations in those with B-ALL associated with familial predisposition variants compared to sporadic cases.

cancer biology↗

bioRxiv2026

Single-cell multiomic mapping of genetic predisposition to childhood B-cell acute lymphoblastic leukemia

Inherited genetic variation substantially increases the risk for developing childhood B-cell acute lymphoblastic leukemia (B-ALL), the most common cancer in children, yet the underlying mechanisms remain poorly understood. To address this limitation, we employ a single-cell multiomic framework to functionally dissect common regulatory variants associated with B-ALL risk. Coupling this multiomic analysis with assessment of allelic skews in chromatin accessibility, we reveal the impact of risk alleles and disruptions in transcription factor networks specific to B-cell progenitors, thereby providing mechanistic insights into altered regulatory programs underlying B-ALL predisposition. By constructing long-range variant-to-target gene maps, we identify 34 high-confidence B-ALL susceptibility genes. Among these, we uncover and functionally validate a risk allele that selectively upregulates expression of ELK3, a previously unrecognized regulator of B-cell development and leukemogenesis. Together, these findings establish a comprehensive variant-to-function map of cell state-specific regulatory disruptions underlying inherited predisposition to B-ALL and define new risk mechanisms, which could pave the way for future targeted prevention approaches.

cancer biology↗

bioRxiv2026

Breast cancer ovarian metastases show increased activity of GPCR pathways

Treatment resistance and metastases occur in 10-20% of patients with invasive lobular carcinoma (ILC), the most common special histological subtype of breast cancer. ILC metastasizes to the ovary more frequently than no special type (NST) tumors, also known as invasive ductal carcinoma (IDC). To characterize the genomic landscape of breast cancer ovarian metastases, we analyzed 15,613 local breast cancers, 22,010 non-ovarian metastases, and 246 ovarian metastases sequenced using FoundationOne(R)CDx or FoundationOne(R) assays. Ovarian metastases had enriched CDH1, PIK3CA, and TBX3 mutations and depleted TP53 and MYC alterations relative to local breast cancers, with additional depletion of ESR1 mutations compared to non-ovarian metastases. CDH1 mutations were less frequent in ovarian metastases (47%) than local ILC (81.3%), with reduced 16q loss (64% vs 84%), indicating that ovarian metastases also arise from non-ILC tumors. We extended these findings to a UPMC cohort of 27 ovarian metastases (13 ILC, 8 IDC, 6 mixed ductal-lobular carcinoma) with patient-matched primary tumors in most cases. In both cohorts, patients with ovarian metastases were significantly younger than those with other metastatic sites. In the UPMC cohort, the most frequent mutations were in PIK3CA, CDH1, KMT2C, FOXA1, and RUNX1. Transcriptomic analysis identified upregulated G protein-coupled receptor (GPCR) pathways, including metabotropic glutamate receptor signaling. Functional studies showed that calcium-sensing receptor (CaSR), a GPCR overexpressed in ovarian metastases, drives MEK/ERK-dependent migration and F-actin reorganization in ILC cell lines, enhanced by estrogen and blocked by calcilytic, MEK, or anti- estrogen treatment. Our findings inform future therapeutic targeting of ovarian metastasis.

cancer biology↗

bioRxiv2026

Molecular principles for graded activation of dopamine D1 receptor

G protein-coupled receptors (GPCRs) signal across a continuum of activation states, yet how ligands encode distinct signaling efficacies remains poorly understood. Here, we define the molecular mechanism underlying graded activation of the dopamine D1 receptor (D1R), a major therapeutic target for neuropsychiatric disorders. Functional analyses reveal that two widely used pharmacological tools, LE300 and SCH23390, possess intrinsic efficacy as an inverse agonist and a weak partial agonist, respectively, rather than the efficacy-silent neutral antagonists. Structural, molecular dynamics and mutagenesis analyses capture previously unrecognized inactive and intermediate receptor activation states that bridge known active conformations, and reveal that ligand efficacy is encoded through the progressive engagement of a conserved activation pathway centered on the W6.48 toggle switch. Guided by this mechanism, a single chemical modification markedly increases the agonist efficacy of SCH23390. Comparison with dopamine D2 receptor structures further reveals a conserved mechanism of inverse agonism despite distinct subtype-specific recognition. Together, these findings establish a structural framework for graded agonism at D1R and provide general principles for the rational design of efficacy-tuned therapeutics at dopamine receptors and related GPCRs.

biochemistry↗

bioRxiv2026

The SoxC transcription factors regulate multiple early retinal lineages and function in parallel with Atoh7 to promote retinal ganglion cell genesis

Retinal development is orchestrated by a network of transcription factors that guide multipotent retinal progenitor cells (RPCs) to fate-committed lineages, ultimately producing seven major retinal cell classes. Among these, retinal ganglion cells (RGCs) serve as the sole output neurons of the retina, relaying visual and non-visual information to the brain. RGC specification requires a cascade of transcriptional regulators, including the basic helix-loop-helix (bHLH) factor Atoh7, which confers competence to RPCs, and Pou4f2 and Isl1, which drive terminal differentiation and subtype diversification. Previous studies demonstrate that the SoxC group transcription factors (Sox4, Sox11, and Sox12) are also involved in RGC genesis, but their precise integration into the Atoh7-driven regulatory hierarchy remains undefined. To address this question, we used the retina-specific Vsx2-Cre line to generate Sox4/Sox11 double conditional knockout (dcKO) and Sox4/Sox11/Atoh7 triple knockout (tKO) mice. Immunohistochemistry revealed profound lineage disruption and reduced progenitor proliferation and survival in both dcKO and tKO retinas; not only RGC genesis but also that of horizontal and amacrine (H&As) cells were severely compromised, whereas photoreceptor cells (PHCs) production increased. These results indicate that Sox4 and Sox11 are involved in the coordinated generation of the different early retinal lineages. Like the Atoh7-null retina, RGC precursors still formed in the SoxC dcKO retina, but their genesis was almost completely abolished in the tKO retina. Our findings indicate that the SoxC factors act in parallel with Atoh7 as a major upstream regulatory input to initiate RGC fate. Bulk RNA-seq revealed the SoxC-dependent transcriptional programs and signaling pathways and confirmed the lineage changes demonstrated by marker analysis. CUT&Tag analysis identified the genome-wide binding sites and thereby the target genes of Sox11, further illuminating the mechanisms underlying functions of the SoxC factors in multiple retinal cell states/types during development.

developmental biology↗

bioRxiv2026

MGM2 as a Unified Foundation Model for Microbiome World Exploration

Microbiomes are information-rich biological systems, yet most computational analyses still reduce communities to cohort-specific abundance tables. Here we introduce MGM2, a multimodal foundation model pretrained on 1,821,291 MicrobeAtlas samples and 225,067 OTUs clustered at 99% sequence similarity. MGM2 couples NTv3-derived microbial sequence embeddings with abundance conditioning and community-semantic alignment to learn transferable sample- and token-level representations. Frozen MGM2 representations outperformed DeepPhylo by 0.06-0.21 macro-AUROC across five temporally held-out MGnify hierarchy levels, with the largest gains for rare and fine-grained labels. In fecal microbiota transplantation, MGM2-XLarge achieved a response ROC AUC of 0.79 and reduced post-transplant Bray-Curtis distance by 15% relative to the recipient baseline. The same representation supported ASV-level trend forecasting across 24 wastewater treatment plants. Sparse autoencoder analysis resolved MGM2-XLarge token states into a 4,096-feature dictionary spanning taxonomic identity, abundance state, ecological context and technical variation. MGM2 therefore provides a sequence-aware and interpretable representation layer for microbiome classification, paired-community prediction, forecasting and feature discovery. Highlights[bullet] MGM2 integrates sequence, abundance and community semantics through pretraining on 1.82 million microbiome samples. [bullet]Frozen MGM2 improved macro-AUROC over DeepPhylo by 0.06-0.21 across five temporally held-out MGnify levels. [bullet]MGM2-XLarge reached a response ROC AUC of 0.79 and reduced post-FMT Bray-Curtis distance by 15%. [bullet]A 4,096-feature sparse autoencoder atlas resolves taxonomic, abundance, ecological and technical signals.

bioinformatics↗

bioRxiv2026

Promoter architecture decodes response regulator phosphorylation into distinct transcriptional logics in Pseudomonas aeruginosa

Two-component systems convert environmental signals into transcriptional responses, but how one response-regulator phosphorylation state produces different outputs at different promoters remains unclear. Here we show that promoter architecture determines how the ParR response regulator in Pseudomonas aeruginosa interprets phosphorylation. Phenotypic, transcriptomic, biochemical and promoter-engineering analyses showed that phosphorylation of the conserved ParR receiver residue Asp57 lowered DNA-occupancy thresholds across target promoters. Individual promoters, however, converted this shared increase in binding into activation, repression or phosphorylation-dependent sign switching. Chromosomally D57 ParR variants reproduced these behaviours at endogenous loci and altered antibiotic susceptibility, biofilm formation and virulence-associated phenotypes. Mapping and engineering of representative promoters identified a two-tier cis-regulatory code: half-site sequence compatibility determines productive ParR engagement, whereas spacer length determines regulatory sign and magnitude. Spacer swaps were sufficient to reprogram promoter logic between regulatory modes. These findings separate regulator state from promoter decoding and show how bacterial promoter architecture can diversify the outputs of a shared phosphorylation signal. ImportancePseudomonas aeruginosa is an opportunistic pathogen that survives antimicrobial treatment and adapts to diverse host environments by rapidly changing gene expression. How a single sensory pathway generates different, and sometimes opposing, responses across many genes has remained unclear. We show that the regulatory protein ParR does not impose one fixed transcriptional program. Instead, the DNA architecture of each target promoter determines whether the same phosphorylation signal activates, represses, or reverses gene expression. The sequence and spacing of ParR recognition sites form a compact regulatory code, and changing spacer length alone can reprogram promoter behavior. Different ParR states were also associated with changes in antibiotic susceptibility, biofilm formation, and virulence-associated phenotypes. Our findings explain how one bacterial signaling pathway can generate diverse physiological outcomes without additional regulators and provide a framework for understanding regulatory flexibility in Pseudomonas aeruginosa and for engineering genetic circuits with multiple outputs. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=142 SRC="FIGDIR/small/740031v1_ufig1.gif" ALT="Figure 1"> View larger version (55K): org.highwire.dtl.DTLVardef@1b6ec28org.highwire.dtl.DTLVardef@4a3f2eorg.highwire.dtl.DTLVardef@79e528org.highwire.dtl.DTLVardef@b3aa77_HPS_FORMAT_FIGEXP M_FIG C_FIG

microbiology↗

bioRxiv2026

Promoter architecture diversifies ParR-dependent transcription in Pseudomonas aeruginosa

Two-component systems convert environmental signals into transcriptional responses, but how one response-regulator phosphorylation state produces different outputs at different promoters remains unclear. Here we show that promoter architecture determines how the ParR response regulator in Pseudomonas aeruginosa interprets phosphorylation. Phenotypic, transcriptomic, biochemical and promoter-engineering analyses showed that phosphorylation of the conserved ParR receiver residue Asp57 lowered DNA-occupancy thresholds across target promoters. Individual promoters, however, converted this shared increase in binding into activation, repression or phosphorylation-dependent sign switching. Chromosomally D57 ParR variants reproduced these behaviours at endogenous loci and altered antibiotic susceptibility, biofilm formation and virulence-associated phenotypes. Mapping and engineering of representative promoters identified a two-tier cis-regulatory code: half-site sequence compatibility determines productive ParR engagement, whereas spacer length determines regulatory sign and magnitude. Spacer swaps were sufficient to reprogram promoter logic between regulatory modes. These findings separate regulator state from promoter decoding and show how bacterial promoter architecture can diversify the outputs of a shared phosphorylation signal. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=142 SRC="FIGDIR/small/738673v1_ufig1.gif" ALT="Figure 1"> View larger version (54K): org.highwire.dtl.DTLVardef@1a0a26forg.highwire.dtl.DTLVardef@10836a8org.highwire.dtl.DTLVardef@3c6cd8org.highwire.dtl.DTLVardef@10dfefb_HPS_FORMAT_FIGEXP M_FIG C_FIG HighlightsParR coordinates antibiotic resistance, biofilm formation and virulence-linked phenotypes in P. aeruginosa. Promoters convert ParR phosphorylation into activation, repression or sign switching. Half-site compatibility and spacer length form a cis-regulatory code for ParR output.

biochemistry↗

bioRxiv2026

MAERM: Predicting Enzyme-Reaction Matching Relationships with a Mixed-Attention Model

Harnessing enzyme specificity requires a thorough understanding of enzyme promiscuity, which determines enzymes catalytic scope; however, measuring this scope still relies heavily on labor-intensive analytical approaches. While data-driven approaches have emerged to predict the catalytic scope of enzymes, these methods continue to face challenges such as restricted datasets and insufficient integration of enzyme structural information and reaction transformations. Here, we introduce MAERM, an innovative mixed-attention model designed to predict enzyme-reaction matching relationships. Built on our MAERM-DB, a dataset with broad coverage of validated and chemoenzymatic catalysis data, MAERM utilizes a local-global attention module to integrate multimodal enzyme information with fine-grained reaction representations, thereby predicting enzyme-reaction matching probabilities. Results show that MAERM consistently outperforms all baselines, with an average F1-score of 0.984. Notably, on challenging test samples with less than 40% sequence identity to the training set, MAERM outperforms the second-ranked model by 5.9% in F1-score. In addition, MAERM achieves the highest top-10 success rate of 51.7% on Enzyme-405 and the highest balanced accuracy of 0.697 on BioCat-547, further supporting its generalizability in enzyme screening and chemoenzymatic catalysis. Finally, MAERM can serve as an efficient scoring module. When integrated with ProteinMPNN, MAERM has successfully guided novel enzyme design for two carbonyl reduction reactions, resulting in enhanced catalytic potential for the native substrate and demonstrating broad compatibility. Overall, MAERM has the potential to reduce the experimental cost of measuring enzymes catalytic scope, facilitate enzyme design, and ultimately accelerate the design-build-test-learn cycle in enzyme engineering.

bioinformatics↗

bioRxiv2026

Therapeutic targeting of MYC- and MYCN-driven medulloblastoma with a novel MYC degrader molecule

BackgroundMedulloblastoma (MB) is the most common malignant brain tumour in children, and aggressive subgroups are frequently driven by the oncoproteins MYC or MYCN. Direct therapeutic targeting of MYC/MYCN has been challenging because of their intrinsically disordered protein structures. The aim of this study was to determine whether novel SE486-11 analogues (UNSW-SCs) can therapeutically target MYC/MYCN-driven MB. MethodsThe anticancer activity of UNSW-SCs was assessed in MB cell lines with differential MYC/MYCN expression. Target engagement was evaluated using surface plasmon resonance and drug affinity responsive target stability assays. Blood-brain barrier penetration, MYC/MYCN protein degradation, cell cycle effects, apoptosis, DNA damage, and synergy with histone deacetylase (HDAC) inhibitors were examined. Therapeutic efficacy was evaluated in murine models of MYC- and MYCN-driven human MB. ResultsUNSW-SCs showed potent anticancer activity, with preferential selectivity toward MB cells expressing high MYC/MYCN levels and IC50 values ranging from 0.22 to 1.18 M. The lead molecule, UNSW-SC-22, directly bound MYC, crossed the blood-brain barrier, and achieved a brain-to-plasma ratio of 1.44 at peak concentrations. UNSW-SC-22 induced MYC/MYCN-dependent cytotoxicity associated with enhanced proteasomal degradation, cell cycle arrest, apoptosis, and DNA damage. Combined treatment with HDAC inhibitors further reduced MYC/MYCN protein levels, increased DNA damage, and enhanced apoptosis. In vivo, UNSW-SC-22, either alone or with entinostat, significantly suppressed intracranial tumour growth and prolonged survival. ConclusionsUNSW-SC-22 is a brain-penetrant MYC/MYCN-targeting molecule with potent preclinical activity in MYC/MYCN-driven MB, supporting its development as a monotherapy or combination strategy with HDAC inhibition. Key PointsO_LIUNSW-SC-22 directly targets MYC/MYCN in medulloblastoma. C_LIO_LIUNSW-SC-22 crosses the blood-brain barrier and prolongs survival. C_LIO_LIHDAC inhibition enhances UNSW-SC-22 activity in MYC-driven cells. C_LI Importance of the StudyMYC- and MYCN-driven medulloblastomas remain among the most aggressive paediatric brain tumours, yet direct pharmacological targeting of MYC/MYCN has historically been difficult. This study identifies UNSW-SC-22 as a novel brain-penetrant small molecule that directly engages MYC/MYCN and promotes proteasomal degradation, leading to tumour cell death. Compared with prior MYC-directed approaches, UNSW-SC-22 combines direct target engagement, blood-brain barrier penetration, and in vivo efficacy in orthotopic MYC- and MYCN-driven medulloblastoma models. The study further demonstrates that combining UNSW-SC-22 with histone deacetylase inhibition enhances MYC/MYCN suppression, DNA damage, apoptosis, and survival benefit. These findings provide a translational framework for developing MYC/MYCN degradation as a therapeutic strategy for high-risk medulloblastoma and support future pharmacological optimisation, biomarker-guided patient selection, and combination therapy development.

cancer biology↗

bioRxiv2026

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

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

pathology↗

bioRxiv2026

A body-brain-behavioral cross-frequency architecture links biological and behavioral periodicities

Spontaneous behaviors, e.g., walking and speaking, are thought to rely on an internal sense of time that provides a scaffold for precise temporal coordination. Yet the endogenous rates of many behavioral processes often diverge from the arbitrarily defined unit of objective time (Chronos), raising a fundamental question: what temporal reference frame coordinates behavior? Recent theoretical work (Buzsaki, 2026) proposed the rich repertoire of subjective time (Kairos) to fluctuate in function of a dynamic cross-frequency architecture linking body-brain periodicities along a lognormal linear progression in the frequency domain. Withing this framework, an emergent sense of time may regulate the rate of semi-periodic behaviors. Using high temporal resolution multimodal recordings, we show that endogenous body-brain periodicities - including pupil fluctuations, saccadic eye movements, respiration, cardiac activity and neural oscillations - as well as spontaneous behaviors such as tapping, walking, and speaking, are organized as an arithmetic progression in a natural logarithmic frequency space. This observation suggests that a unified scaling law may coordinate complex, multi-scale interactions across biological and behavioral timescales. We propose that such organization may provide an emergent temporal reference frame that scaffolds perception and action.

neuroscience↗

bioRxiv2026

Super-Resolution Visual ProteomEx for Hard Tissues and Clinical Samples

Correlating nanoscale tissue architecture with unbiased molecular composition within the same specimen remains a fundamental challenge in spatial biology. Current spatial proteomics methods either lack the imaging resolution required to resolve nanoscale tissue architecture or achieve only limited proteome coverage, and untargeted approaches capable of deep protein identification have yet to be integrated with super-resolution imaging within a single workflow. Here, we present microProteomEx, an integrated platform that combines hydrogel-assisted tissue expansion with mass spectrometry-compatible fluorescence imaging, laser capture microdissection, and bottom-up proteomics to achieve simultaneous three-dimensional super-resolution imaging (effective lateral resolution down to [~]47 nm on conventional diffraction-limited microscopes) and spatially resolved proteomics at a scalable lateral resolution of 29-100 {micro}m (0.02-0.28 nL volumetric resolution). By optimizing fixation, protein anchoring, and a secondary re-embedding strategy, we extend the method to mechanically resilient tissues, including mouse kidney and heart, as well as to formalin-fixed, paraffin-embedded clinical specimens. We demonstrate single-glomerulus and single-plaque proteomics (450-1,000 proteins per structure), resolve proteomic differences between malignant melanoma and giant congenital melanocytic nevus in a rare pediatric case, and characterize the morphology-resolved molecular heterogeneity of cored versus diffuse amyloid-beta plaques in a mouse model of Alzheimers disease across two disease stages. microProteomEx establishes a broadly accessible framework for correlating tissue ultrastructure with deep spatial proteomics, with direct implications for disease biology, biomarker discovery, and precision medicine.

biochemistry↗

bioRxiv2026

In vivo gene disruption and homology-directed repair in muscles and muscle stem cells using CRISPR/Cas9

Programmable endonucleases such as CRISPR/Cas9 provide powerful tools to edit mammalian genomes by engaging cellular mechanisms of DNA double-strand break (DSB) repair. CRISPR-catalysed homology-directed repair (CRISPR-HDR), though generally less efficient than other modes of DNA repair, holds particular promise to enable precise sequence replacement by targeted insertion of a homologous DNA template1,2. While recent studies have reported appreciable levels of HDR in cardiomyocytes in vivo3, skeletal muscle myofibres have historically been considered refractory to HDR-mediated genome editing4. Furthermore, how repair outcomes differ across tissues after systemic delivery of CRISPR/Cas9 editors, whether precise HDR editing can be achieved in regenerative tissue stem cells, and how developmental timing influences accessibility to CRISPR-induced repair remain unclear. Here, we use an adeno-associated virus (AAV)-delivered in vivo GFP-to-BFP colour-switching reporter system (AAV-GFP-to-BFP) to examine in vivo CRISPR-HDR with cellular- and tissue-level resolution. We find that postnatal cardiac muscle, skeletal muscle, and muscle stem cells undergo templated HDR at different rates across discrete developmental stages in mice. While HDR-edited muscle stem cells and myofibres were readily detectable after in vivo editing in juvenile mice, editing in neonatal mice yielded more efficient HDR in cardiac tissue. Based on these results, we adapted the CRISPR-HDR approach to rescue the therapeutically relevant Dmd mutation in mdx mice, demonstrating recoding to the wild-type protein sequence in both skeletal and cardiac muscles. These results provide a framework for advancing donor-templated DNA repair in living postnatal animals, and reveal unexpected cellular, developmental, and disease-related constraints on precise, therapeutic in vivo gene correction.

cell biology↗

bioRxiv2026

Learning Perturbation Effects Through Contrastive Alignment of Multimodal Biological Embeddings

Multimodal single-cell perturbation screens offer a scalable approach for characterizing the effects of genetic and chemical interventions on cellular state. However, most existing representation-learning methods are tailored to a single perturbation modality and fail to explicitly incorporate external semantic knowledge, which limits their ability to generalize across datasets and perturbation types. Here, we introduce PertOmni, a CLIP-style multimodal representation-learning framework that aligns transcriptomic perturbation signatures with text-derived embeddings of curated genes and compound descriptions, as well as image-derived embeddings from cell paintings. PertOmni jointly trains a shared transcriptomic encoder and dataset-specific text encoders using a masked contrastive objective that emphasizes within-cell-type discrimination while mitigating confounding effects arising from cell-type heterogeneity. We evaluate the produced joint embedding space on bi-directional retrieval, drug-gene interaction inference, and perturbation prediction across both small-molecule and CRISPRi perturbation datasets, and demonstrate consistent improvements over strong baseline methods.

bioinformatics↗

bioRxiv2026

Systematic benchmarking of zero-shot utility and robustness in single-cell transcriptomic foundation models

Single-cell foundation models (scFMs) have emerged as powerful representation learning approaches for single-cell transcriptomics. However, the utility and robustness of their pretrained representations across diverse analytical tasks and data conditions remain insufficiently characterized, particularly in zero-shot settings without task-specific fine-tuning. Here, we systematically analyze zero-shot performance of single-cell transcriptomic representations across 20 methods, 6 downstream tasks and 1,607 datasets comprising nearly 21.8 million cells. We evaluate model behavior along three complementary dimensions: utility on original datasets, robustness to controlled changes in dataset structure, and exploratory associations between dataset characteristics and performance variation. Our results show that scFM performance is strongly task dependent, with no single method consistently outperforming others across cell- and gene-level analyses. Notably, high utility on original datasets did not necessarily translate into robustness under structural perturbations, and several top-ranking methods were sensitive to changes in cell number, gene number, class composition, class imbalance, and batch complexity. Conventional statistical and task-specific methods remained competitive in several settings, while greater computational cost did not consistently correspond to better performance. Driver analyses further identified task-specific associations between performance and dataset characteristics, including cell-type complexity, train-test class overlap, batch number, and regulatory target-set size. Together, these findings show that the zero-shot utility and robustness of pretrained scFM representations depend jointly on analytical task and dataset structure. Our study provides a practical basis for context-aware representation selection and underscores the importance of evaluating structural robustness alongside utility when developing and applying scFMs.

bioinformatics↗

bioRxiv2026

Detection of Melanoma Using Deep Serum Proteome Profiling and Machine Learning

Early detection determines melanoma outcomes, yet current screening relies on visual inspection that misses molecular changes preceding clinical diagnosis. To identify serum protein signatures of melanoma development, we leveraged the Department of Defense Serum Repository, analyzing 390 longitudinal serum samples from 73 melanoma cases and matched controls across four timepoints (4 and 2 years prior, diagnosis, and 2 years post-diagnosis) using data-independent acquisition mass spectrometry with Seer Proteograph nanoparticle enrichment. We quantified 3,364 proteins and applied machine-learning strategies combining cross-sectional case-control comparisons with longitudinal tracking of within-individual changes. Cross-sectional analysis at diagnosis achieved AUC 0.823 (95% CI: 0.723-0.923), identifying eight consensus features selected in >50% of cross-validation folds: PRSS1, GSK3B, FAM20C, CES1, CCL14, EPHA10, LMAN2, and ITIH1. These signatures, spanning proteases, immune activation, and extracellular matrix remodeling, demonstrate proof-of-concept for serum-based melanoma detection and provide candidate biomarkers for validation.

Cancer Biology↗

bioRxiv2026

Antibodies to influenza A virus hemagglutinin and neuraminidase limit egress and alter the physical properties of released virus particles

Influenza A virus (IAV)-specific antibodies neutralize mature virions by inhibiting functional sites or, in some cases, by promoting virion aggregation. Many antibodies also act on infected cells to reduce virion yields, but the underlying mechanisms and effects on the physical properties and function of released virus particles remain incompletely characterized. Here we use flow virometry to acquire high-sensitivity yield and size measurements of virus particles released in the presence of antibodies. Combined with digital-droplet PCR and electron microscopy, this approach enables comprehensive characterization of antibody-induced changes in particle genome-content and morphology. We show that antibodies rapidly and dynamically alter released particle distributions, reducing yields and inducing the production of larger particles. Both effects result in part from aggregation induced by crosslinking of viral antigen on the infected-cell surface and inhibition of viral NA. However, yield reduction is not fully explained by aggregation, and a subset of induced larger particles are elongated virions. Finally, particles formed in the presence of HA stem-binding antibody, which does not inhibit attachment of mature virions, show reduced attachment in subsequent rounds of infection. Altogether, we uncover an unappreciated mechanism by which antibodies interfere with viral infection that occurs only during budding and release. Our work highlights the necessity of studying how the immune response shapes virus populations in the context of active infection processes.

microbiology↗