bioRxiv Science⌕ Search

Biology subjects

Singh, R.

Publications and source records attributed to Singh, R..

At least 55 records · Page 3Linked to original sources
bioRxiv2026

Ancient intraflagellar transport machinery controls unique spatial distribution of phototropin in an evolutionary important non-flagellated vegetative stage of terrestrial alga

Intraflagellar transport (IFT) is a conserved trafficking system in eukaryotes that moves proteins along microtubules. It is best known for its essential role in building and maintaining cilia and flagella. Intriguingly, several IFT components are still found in organisms that no longer possess flagella, raising important questions about their original functions and how they may have been repurposed during evolution. The filamentous alga Klebsormidium nitens, positioned at the base of the streptophyte lineage, offers a valuable model for exploring this transition. Here, we investigate the IFT machinery in K. nitens and its relationship with the blue-light photoreceptor phototropin. Comparative genomic analyses show that key IFT-A and IFT-B components are retained, despite the complete loss of flagella in vegetative state. Cellular detection and immunofluorescence studies revealed the presence and localisation of IFT components, interestingly, their co-localization with phototropin. Notably, IFT-139 and IFT-20 strongly co-localize with phototropin at plasma membrane-associated regions. Phototropin overlapping localization (plasma membrane associated) with conserved phospho-adaptor protein 14-3-3, pointing to a phosphorylation-dependent signaling network. Unlike in Chlamydomonas reinhardtii, where these proteins localize to flagella, their interaction in K. nitens occurs independently of cilia presence. Together, these results evidenced that IFT components were retained and repurposed early in streptophyte evolution and might support phototropin localization and signalling, revealing an ancestral, non-ciliary role for the IFT system.

cell biology↗
bioRxiv2026

EvoPool: Evolution-Guided Pooling of Protein Language Model Embeddings

Protein language models (PLMs) encode amino acid sequences into residue-level embeddings that must be pooled into fixed-size representations for downstream protein-level prediction tasks. Although these embeddings implicitly reflect evolutionary constraints, existing pooling strategies operate on single sequences and do not explicitly leverage information from homologous sequences or multiple sequence alignments. We introduce EvoPool, a self-supervised pooling framework that integrates evolutionary information from homologs directly into aggregated PLM representations using optimal transport. Our method constructs a fixed-size evolutionary anchor from an arbitrary number of homologous sequences and uses sliced Wasserstein distances to derive query protein embeddings that are geometrically informed by homologous sequence embeddings. Experiments across multiple state-of-the-art PLM families on the ProteinGym benchmark show that EvoPool consistently outperforms standard pooling baselines for variant effect prediction, demonstrating that explicit evolutionary guidance substantially enhances the functional utility of PLM representations. Our implementation code is available at https://github.com/navid-naderi/EvoPool.

bioinformatics↗
bioRxiv2026

Deciphering the Therapeutic Potential of Resveratrol Against Pancreatic Cancer Through Network Pharmacology

Resveratrol, a naturally occurring polyphenol, exhibits anticancer, anti-inflammatory, and antioxidant properties. However, its molecular mechanisms in pancreatic cancer remain incompletely understood, necessitating integrative approaches such as network pharmacology and molecular docking. Potential resveratrol targets were identified using Swiss Target Prediction, while pancreatic cancer-related genes were retrieved from GeneCards and NCBI Gene databases. Overlapping targets were obtained through Venn analysis, followed by protein- protein interaction network construction, hub gene selection, and enrichment analysis. Molecular docking validated compound-target interactions. A total of 100 predicted resveratrol targets and 1,447 pancreatic cancer- associated genes were screened, yielding 39 overlapping genes. Network analysis identified hub genes including EGFR, SRC, MTOR, PIK3CA, PIK3CB, BCL2, and PTGS2. Gene Ontology enrichment indicated roles in cell proliferation, apoptosis regulation, inflammatory response and metabolic regulation while KEGG pathway analysis highlighted the PI3K-Akt, ErbB, and EGFR inhibitor resistance signaling as being closely associated with pancreatic cancer pathway. Docking analysis revealed strong binding of resveratrol with KRAS (-8.2 kcal/mol), EGFR (-7.9 kcal/mol), and MTOR (-7.7 kcal/mol), stabilized by hydrogen bonding. The interaction with KRAS, although not among the predicted targets. Expression profiling validated upregulation of hub genes in tumor samples. Resveratrol exerts multi-targeted effects in pancreatic cancer by modulating oncogenic pathways, particularly KRAS and PI3K/Akt/mTOR signaling. Its favourable safety profile and robust hub gene interactions highlight its potential as a supplementary therapeutic drug, necessitating further preclinical and clinical confirmation.

bioinformatics↗
bioRxiv2026

The mood stabilizer lithium alters behaviour and physiology via the gut brain axis.

Lithium, introduced 75 years ago by John Cade1, remains the most effective mood stabilizer for bipolar disorder2. Lithium is proposed to modulate an array of cellular pathways, many ubiquitous to all cells, with pleiotropic roles unlinked to bipolar disorder or lithium responsiveness in genome-wide association studies3,4. These mechanisms cannot explain lithiums specific effects on mood and behaviour. We demonstrate that lithiums primary action is in the periphery, not in the brain itself. Lithium acts in the gut to trigger behavioural and physiological changes, akin to those associated with a torpor-like state, that protect individuals from ingested toxins. Lithium activates gastrointestinal enterochromaffin (EC) cells via their Trpm2 cation channels to modulate afferent vagal and area postrema inputs to the brain. Eliminating these inputs by focal brain lesions eliminates lithiums effects, as does ablation of EC cells or their Trpm2 expression. Lithiums Trpm2-dependent activation of EC cells also occurs in human gut tissue, providing translational relevance for our discovery. These findings challenge the prevailing perception that lithium acts directly on the brain. Via a previously unsuspected gut-brain pathway, lithium engages brain circuitry that reduces arousal and interaction with the external world, therapeutic goals in the manic phase of bipolar disorder.

neuroscience↗
bioRxiv2026

DeltaNMF: A Two-Stage Neural NMF for Differential Gene Program Discovery

Non-negative matrix factorization (NMF) is a foundational dimensionality-reduction method in single-cell transcriptomics, valued for its interpretable gene programs. However, in case-control settings common in perturbation and disease research, standard NMF conflates quantitative shifts in program usage with qualitative emergence of novel programs, obscuring whether observed differences reflect altered activity of shared programs or genuinely new cellular states. Here, we present DeltaNMF, a neural network-based reformulation of NMF that enables flexible biological priors and structured multi-stage fitting. We first demonstrate how our framework incorporates foundation model-derived gene similarities through graph-Laplacian regularization, improving program coherence by 2.4-fold on protein interaction networks. Building on this flexibility, we introduce a two-stage architecture that explicitly separates baseline programs (learned from control cells) from case-specific programs, disambiguating altered usage from novel program emergence. Our GPU-accelerated implementation achieves 20x speedup over consensus NMF while maintaining comparable accuracy. On synthetic data with known ground truth, DeltaNMF correctly identifies which programs are new versus shifted. Applied to coronary artery disease (CAD) Perturb-seq data, DeltaNMF isolates disease programs including the cerebral cavernous malformations pathway identified by Schnitzler et al., while clearly distinguishing them from baseline endothelial programs with altered activity. Our neural network formulation of NMF opens new directions for incorporating diverse biological priors into interpretable single-cell analysis, providing a principled framework for differential program discovery in case-control studies.

bioinformatics↗
bioRxiv2026

Host-Botrytis co-transcriptomics reveals finely tuned interactions with closely related legumes

Generalist pathogens infect diverse plant hosts, yet how these interactions differ across hosts is poorly understood. Here, we conduct a molecular analysis of a generalist pathogen interacting with closely related hosts. A co-transcriptomic framework is used to dissect host-pathogen interactions between the generalist necrotroph Botrytis cinerea and two closely related legume hosts, common bean (Phaseolus vulgaris) and cowpea (Vigna unguiculata). Using a diverse set of 72 Botrytis isolates, we quantified lesion development alongside host and pathogen gene expression. Although lesion formation was driven primarily by pathogen genetic variation, transcriptomic responses in both host and pathogen exhibited significant host x isolate interactions. This indicated that extensive, fine-scale transcriptional plasticity created similar disease outcomes. Botrytis genes showing host-specific expression were enriched for cell wall-modifying enzymes and some specialized metabolic genes, indicating greater host responsiveness of these core virulence mechanisms than previously appreciated. Co-expression network analysis in both host and pathogen further showed that in both organisms, gene membership for individual networks are restructured in response to genetic diversity. For example in Botrytis, we identify different sets of genes host-dependently co-expressing with a non-ribosomal peptide synthetase (NRPS) gene cluster, suggesting divergent functional deployment of the same virulence machinery across closely related hosts. Both legume species exhibited extensive isolate-dependent transcriptional reprogramming, with approximately two-thirds of expressed host genes responding to pathogen diversity. While conserved defense pathways such as jasmonate/ethylene signaling and phenylpropanoid metabolism were upregulated in both hosts, the specific genes in the networks differed markedly, highlighting lineage-specific rewiring of defense strategies. These results suggest that generalist pathogen success is underpinned by pervasive gene expression plasticity in both host and pathogen, allowing similar phenotypic outcomes to emerge from highly divergent molecular states. SummaryO_LIGeneralist pathogens infect diverse plant hosts, yet how these interactions differ across hosts is poorly understood. This study investigates how a generalist pathogen achieves successful infection across closely related hosts, and how these hosts respond. C_LIO_LIA co-transcriptomic approach was applied to interactions between 72 genetically diverse isolates of the fungal necrotroph Botrytis cinerea and two legume hosts, common bean and cowpea. Lesion development and host and pathogen gene expression were quantified. C_LIO_LILesion formation was primarily driven by pathogen genetic variation, yet both host and pathogen transcriptomes showed strong host x isolate interactions. Both host and pathogen balance conserved responses with finely tuned, host-specific mechanisms. Further, host-dependent transcriptional responses involve network modulation around a common core of genes in both host and pathogen. C_LIO_LIGeneralist pathogen success is underpinned by pervasive gene expression plasticity in both host and pathogen, allowing similar phenotypic outcomes to emerge from highly divergent molecular states. C_LI

plant biology↗
bioRxiv2026

A multi-plant transcriptomic atlas reveals conserved and lineage specific defense architectures in response to Botrytis cinerea

Generalist pathogens pose a challenge to plant immunity by infecting diverse hosts while harboring extensive intraspecific genetic variation. Whether evolutionary distant plant lineages rely on a shared immune strategy or deploy distinct, lineage-specific defenses when confronted by these genetically variable members of the same pathogen species remains unresolved. Here, we employed a large-scale co-transcriptomic approach to map the immune landscape of ten diverse eudicot species infected with 72 genetically distinct Botrytis cinerea isolates. We identified a limited core of evolutionarily conserved defense orthologs, along with a vast landscape of lineage-specific transcriptional rewiring. While the broad physiological outcome such as metabolic reprogramming, cell wall modification, and suppression of growth-associated processes was shared across hosts, the regulatory pathways governing this were largely lineage-specific. Crucially, this immune landscape is dynamically shaped by pathogen diversity. Nearly three-quarters of host transcriptional responses were isolate-dependent, with the magnitude of defense activation defined by specific host-isolate combinations rather than a universal species-level response. Even host responses to shared virulence factors, including broadly expressed pathogens phytotoxins, were lineage specific. These findings show that plant immunity to generalist pathogens is built on conserved physiological outcomes executed through rapidly evolving, lineage-specific regulatory programs. This distinct regulatory architecture creates an immune landscape heavily modulated by specific host-isolate combinations, highlighting the necessity of integrating pathogen diversity into models of plant defense evolution and resistance breeding. Significance StatementAchieving durable, broad-spectrum crop protection remains difficult because plant immunity models often rely on limited species and overlook natural genetic diversity. Effective crop protection requires understanding how defense networks operate across diverse lineages. We tested this by measuring immune responses of ten phylogenetically diverse eudicots infected with 72 genetically distinct Botrytis cinerea isolates. We found that plants share conserved physiological defense outcomes achieved through highly divergent, lineage-specific regulatory networks. Host responses were strongly shaped by pathogen genetic diversity, with identical isolates eliciting different transcriptional responses in different hosts. This demonstrates that plant immunity quantitatively senses pathogen variation and disease outcomes emerge from specific host-isolate combinations. These findings explain the limits of resistance transfer and inform strategies for durable disease control.

genomics↗
bioRxiv2026

A dynamic displacement mechanism drives protein import into mitochondria

Most mitochondrial proteins are produced in the cytosol and imported through the translocase of the outer mitochondrial membrane (TOM) to reach their final destination. Although this protein entry gate has been structurally characterized, it remains unclear how precursor proteins are handed off from the cytosolic receptor domains to the translocation pore. Here we show that the cytosolic domain of Tom22--traditionally viewed as the central TOM receptor--acts not as a structured scaffold but as a largely disordered, flexible segment that plays an active role in precursor transfer. Atomic-level structural techniques and in vivo experiments identified a conserved short linear motif that forms a transient !-helical element within this disordered domain. By binding to the canonical precursor protein binding sites of the receptors Tom20 and Tom70, this critical -helical segment acts as a precursor protein displacement element (PPDE). This competitive interaction facilitates the release of preproteins directly above the import pore, and thereby drives translocation across the outer mitochondrial membrane. These findings reveal that flexibility, rather than rigid structure, underlies the central transfer step of mitochondrial outermembrane protein translocation. Our results point to a versatile mechanism for ligand displacement in chaperone, receptor, and transport systems that must balance selective binding with efficient release.

cell biology↗
bioRxiv2026

Recombinant Photo-tagging Enables Fluorescent Labelling of Biomolecules and Visualization of Liquid-Liquid Phase Separation

Liquid-liquid phase separation (LLPS) of biomolecules has emerged as a fundamental principle governing biological systems. A key aspect to understand LLPS is the ability to visualize these coacervates, which requires labelling strategies. Fluorescent labelling of biomolecules remains one of the widely adopted techniques involving either small-molecule chemical conjugation or recombinant expression alongside a fluorescent protein. These small-molecule fluorophores are typically employed in excess, which further necessitates additional downstream processing steps, including separation of the conjugated molecules from the unreacted fluorophores. On the other hand, recombinantly appended fluorescent protein affects the functioning of the labelled biomolecule owing to its large size. There is still a need to develop a labelling technique with small molecular weight fluorophores that is also genetically encodable. We previously discovered that the C-terminal peptide fragment (CTPF) released during photo-exposure of a green-to-red photoconvertible fluorescent protein (PhoCl1) displayed red fluorescence. Notably, this chromophore displays broad excitation and emission wavelength profiles, rendering it robust under multiple excitation wavelengths. We leverage this phenomenon to achieve post-expression fluorescent labelling by genetically fusing PhoCl1 to the target peptide or protein to be labelled (TPTL). We successfully demonstrated this approach of conferring fluorescence to silica binding peptide, riboflavin kinase enzyme, TEV protease, Lanthanide Binding Peptide and maltose binding protein, all of which exhibited red fluorescence upon photocleavage. Finally, we demonstrate the applicability of this probe to visualize LLPS of CTPF tagged elastin-like polypeptide in the presence of polymeric crowder. We anticipate that this strategy for inducing red fluorescence in non-fluorescent biomolecules via a photocleavable protein will open new avenues for minimally perturbative fluorescent labelling.

bioengineering↗
bioRxiv2026

Learning a PRECISE language for small-molecule binding

Virtual screening of billion-scale compound libraries has become feasible through machine learning approaches. In particular, CoNCISE (RECOMB 2025) introduced drug quantization via code-books, achieving highly scalable and accurate binary predictions. However, drug discovery requires understanding not just whether molecules bind, but where they bind and how to target specific sites. Here, we present PRECISE which leverages CoNCISEs quantized small-molecule representations while operating on the targets 3D structure as its input. The key innovation of PRECISE is reimagining drug-target interaction as compatibility between quantized drug embeddings and a latent representation of the targets surface mesh, enriched with electrostatic and geometric features. PRECISE designs a novel surface representation, interpreted through a geometric deep learning architecture, enabling it to identify binding sites more accurately than state-of-the-art methods (DiffDock-L, Chai, and Boltz-2) while the codebook ensures billion-scale screening capability. Our formulation unlocks zero-shot generalization to complex targets such as metalloproteins and multi-chain complexes. To enable efficient integration with downstream docking workflows, we introduce PO_SCPLOWRECISEC_SCPLOW-MCTS, which combines fast PO_SCPLOWRECISEC_SCPLOW-based screening with selective Vina docking through an iterative Monte Carlo Tree Search approach. By providing both mechanistic understanding and massive scalability, PRECISE delivers capabilities that were previously mutually exclusive in virtual screening.

bioinformatics↗
bioRxiv2025

Selective Activation of NaV1.3 Restores Lymphatic Contractility in Age and Injury

BackgroundIntrinsic lymphatic contractility is essential for tissue fluid balance, immunity and organ function, yet no FDA-approved pharmacologic treatments specifically restore lymphatic contractility. Lymph is returned to the circulation by ion channel-driven cyclic contractions of collecting lymphatic vessels. Although voltage-gated sodium (NaV) channels drive cardiomyocyte excitability, their role in lymphatic muscle cell (LMC) physiology is not well defined. We identified NaV1.3, a NaV channel historically viewed as developmentally restricted and limited in adult tissues, as unexpectedly and selectively expressed in adult lymphatic muscle but absent from heart, vascular smooth muscle, and mature brain. We tested whether selective NaV1.3 activation restores impaired lymphatic pumping in aging and radiation injury. MethodsNaV1.3 expression in LMCs was confirmed through single-cell RNA sequencing analysis and immunostaining of mouse and human lymphatic vessels. Lymphatic contractility was quantified by in vivo fluorescence lymphangiography and interstitial fluid clearance was measured with a new bioluminescence assay. NaV1.3 function was assessed in young, aged, and radiation-injured mice. NaV1.3 knockout (Scn3a-/-) mice established the requirement of NaV1.3 for basal lymphatic excitability and responsiveness to the NaV1.3-specific activator, Tf2. ResultsIn mouse and human lymphatic vessels, NaV1.3 is expressed in adult LMCs. Although dispensable for basal lymphatic contractions, NaV1.3 acted as a pharmacologically recruitable reserve that amplified contractile output. Acute NaV1.3 activation with Tf2 increased lymphangion ejection fraction and accelerated interstitial fluid clearance. Tf2 fully restored lymphatic pumping in aged mice and partially rescued radiation-induced contractile deficits. All Tf2 responses were abolished in Scn3a-/- mice, confirming NaV1.3 dependence. ConclusionsNaV1.3 is a selectively druggable ion channel in adult lymphatic muscle that can be recruited to restore lymphatic pump function across aging and injury. Targeted NaV1.3 activation provides a molecular entry point for treating diseases characterized by lymphatic pump failure, a domain with no existing pharmacologic therapies.

physiology↗
bioRxiv2025

Lipid droplet-sphingolipid crosstalk regulates Doa1 activity and DNA damage response

Neutral lipid droplets (nLDs) are dynamic storage organelles that protect the cells from lipotoxicity. They store excess neutral lipids, sequester heavy metals and serve as membrane reservoirs. However, their role in genome integrity remains unclear. Previous studies have linked the shortened lifespan of LD deficient cells to impaired mitochondrial functions caused by defective sterol metabolism, but the molecular origin of this dysfunction remains elusive. In this study, we demonstrate that lack of nLDs inhibits DNA damage sensing mechanism, causing insensitivity to DNA damaging agents. Despite sustained DNA breaks, LD deficient cells unable to activate the checkpoint kinases or arrest cell cycle progression. Persistent insensitivity to double strand breaks results in accumulation of mutations, loss of nuclear as well as mitochondrial DNA integrity leading to accelerated cellular ageing. Through systematic genetic dissection of nLD synthesis genes, we identify a crucial role of lanosterol esterification in lipid droplet formation and their role in the regulation of DNA damage sensing and repair responses. Our study identifies a key regulatory protein, Doa1 that functionally links lipid droplet biosynthesis to DNA damage repair kinase signalling. Together, these findings establish a previously uncharacterized role of neutral lipid droplets in the maintaining the genome stability by acting not only as metabolic buffers but also as signalling platforms that coordinate DNA damage sensing and repair processes. Our work reveals a new mechanistic connection between cellular lipid homeostasis and genome maintenance, highlighting lipid droplets as critical determinants of healthy cellular aging.

cell biology↗
bioRxiv2025

The protein turnover and trafficking of Chlamyopsin6 is regulated by IFT88 and IFT52 in the Chlamydomonas reinhardtii

Microbial rhodopsin-based optogenetics has been widely applied to diverse mammalian and plant cell types for controlling membrane potential mediated responses. However, trafficking of optogenetically active protein to the desired subcellular organelle is still a major concern in optogenetic field. This could be resolved by studying the trafficking mechanism of optogenetically active protein in the native system. Current study is focused on the trafficking of two of the microbial rhodopsins named Chalmyopsin5 and Chlamyopsin6 in a green alga, Chlamydomonas reinhardtii. Chlamyopsin5 and Chlamyopsin6 are modular in nature and possess rhodopsin, histidine kinase, response regulator and cyclase domain in tandem. Immunolocalization of Chlamyopsin5 and Chlamyopsin6 in wild strain suggests their different subcellular localization; Chlamyopsin5 in eyespot and Chlamyopsin6 in flagella. Extensive immunocytochemistry of Chlamyopsin5 and Chlamyopsin6 was performed in different intraflagellar transport (IFT) components-defective strains of Chlamydomonas to dissect their trafficking mode to the destined subcellular compartment. Our results indicated the trafficking of Chlamyopsin5 to the eyespot to be independent of IFT machinery while Chlamyopsin6 to the flagella to be IFT dependent. Further, we demonstrate that IFT88 and IFT52 stabilizes Chlamyopsin6 and IFT20 interacts with Chlamyopsin6 in Chlamydomonas. Protein interactome of Chlamyopsin5 and Chlamyopsin6 indicate their role in nitrogen assimilation, gametogenesis and photoprotection in co-ordination with other photoreceptors. Collectively, our study enabled us to understand the targeting of Chlamyopsins to the subcellular compartment (eyespot and flagella). This study is important to expand optogenetic application of microbial type modular rhodopsin with histidine kinase and response regular. Further research in this direction is required to resolve the current challenge of targeting of optogenetic protein to desired subcellular compartment. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=167 SRC="FIGDIR/small/693822v2_ufig1.gif" ALT="Figure 1"> View larger version (42K): org.highwire.dtl.DTLVardef@2691edorg.highwire.dtl.DTLVardef@367a61org.highwire.dtl.DTLVardef@9c85c6org.highwire.dtl.DTLVardef@128f9b_HPS_FORMAT_FIGEXP M_FIG C_FIG

cell biology↗
bioRxiv2025

Widespread sex-biased gene expression reflects female-biased longevity in a species with environmental sex determination

Sexes frequently differ in life history traits including body size, lifespan, and age at sexual maturity. Aging, the progressive decline in physiological function and cellular resilience over time, is a central process contributing to sex-specific life histories, yet the mechanisms driving sex differences in aging remain largely unresolved. Long-term mark-recapture efforts revealed a striking pattern of female-biased longevity in the painted turtle (Chrysemys picta), a species with temperature-dependent sex determination. As a result, this species provides a compelling system to examine the mechanisms of sex-specific aging in the absence of sex chromosomes. Here, we characterize sex- and age-associated patterns in the blood transcriptomes of wild painted turtles (n = 93). We identified widespread gene expression differences between females and males (2,347 genes; 13.4% of all filtered genes). In contrast, only six genes showed significant linear relationships with continuous age in both sexes. We also employed a machine learning approach which identified distinct sets of genes for which expression was predictive of age in each sex. Age-related gene expression patterns highlight both conserved molecular pathways with known roles in aging as well as novel gene targets. These findings suggest sex-specific molecular processes underlie sex-biased demographic aging and raise questions regarding the environmental and developmental drivers of sex-biased gene expression.

ecology↗
bioRxiv2025

From Circles to Signals: Representation Learning on Ultra-Long Extrachromosomal Circular DNA

Extrachromosomal circular DNA (eccDNA) is a covalently closed circular DNA molecule that plays an important role in cancer biology. Genomic foundation models have recently emerged as a powerful direction for DNA sequence modeling, enabling the direct prediction of biologically relevant properties from DNA sequences. Although recent genomic foundation models have shown strong performance on general DNA sequence modeling, their application to eccDNA remains limited: existing approaches either rely on computationally expensive attention mechanisms or truncate ultra-long sequences into kilobase fragments, thereby disrupting long-range continuity and ignoring the molecules circular topology. To overcome these problems, we introduce eccDNAMamba, a bidirectional state space model (SSM) built upon the Mamba-2 framework, which scales linearly with input sequence length and enables scalable modeling of ultra-long eccDNA sequences. eccDNAMamba further incorporates a circular augmentation strategy to preserve the intrinsic circular topology of eccDNA. Comprehensive evaluations against state-of-the-art genomic foundation models demonstrate that eccDNAMamba achieves superior performance on ultra-long sequences across multiple task settings, such as cancer versus healthy eccDNA discrimination and eccDNA copy-number level prediction. Moreover, the Integrated Gradient (IG) based model explanation indicates that eccDNAMamba focuses on biologically meaningful regulatory elements and can uncover key sequence patterns in cancer-derived eccDNAs. Overall, these results demonstrate that eccDNAMamba effectively models ultra-long eccDNA sequences by leveraging their unique circular topology and regulatory architecture, bridging a critical gap in sequence analysis. Our codes and datasets are available at https://github.com/zzq1zh/eccDNAMamba.

bioinformatics↗
bioRxiv2025

Effects of a novel proteasome inhibitor, UR238 on the tumor immune microenvironment and growth in epithelial ovarian cancer

High-grade serous ovarian cancer (HGSOC) is a deadly gynecologic malignancy, often diagnosed at an advanced stage and most patients will experience recurrence and resistance to platinum-based chemotherapy. While there are few targeted therapies available for HGSOC, there are no effective immunotherapies available to treat this disease. The ubiquitin-proteasome system (UPS) maintains cellular protein homeostasis by degrading misfolded or damaged proteins. In epithelial ovarian cancer (EOC) elevated expression of proteasome subunit PSMB4 correlates with epithelial ovarian cancer growth and poor prognosis. A proteasome inhibitor has yet to be approved for EOC treatment despite evidence of activity in phase-1/2 clinical trials. Limitations of past generation proteosome inhibitors include sub-optimal solid tumor/tissue penetration due to boronic acid functionality, and poor solubility. Here, we show that a novel proteasome inhibitor, UR238, suppresses viability of several EOC cell lines and reduces tumor burden in both murine xenograft and rat syngeneic models. In a rat EOC model, UR238 treatment reduced tumor burden, shifted the predominantly suppressive immune microenvironment to an inflammatory phenotype with reduced suppressive macrophages, and decreased PD-L1 expression on myeloid cells. Additionally, in a mouse model, UR238 also induced robust immune changes in HGSOC bearing mice without a corresponding reduction in tumor burden. Our data highlights the promise of UR238 for the treatment of ovarian cancers as a single-agent or in combination with immunotherapies, where modulating the immune composition of these tumors will be important in improving outcomes for ovarian cancer patients.

immunology↗
bioRxiv2025

Transforming Biological Foundation Model Representations for Out-of-Distribution Data

Foundation models pre-trained on certain biological data modalities exhibit systematic representational biases when encountering out-of-distribution (OOD) data from new assays. The embedding drift largely arises from instrumentation and protocol-related artifacts rather than true biological variation in cell states or tissue morphology. These drifts are distinct from conventional batch effects and cannot be remedied by retraining as sample sizes are often insufficient, and modifying existing embeddings breaks downstream tools that depend on stable representations. We introduce USHER, an adaptable framework to learn simple transforms that return OOD embeddings to a foundation models reference space. USHER enables embedding transformation via an expectation maximization-style procedure. Given a reference in-distribution sample, USHER first estimates a Fused Gromov-Wasserstein coupling that aligns unpaired OOD (source) and reference (target) embeddings by minimizing transport distance while preserving local structure. To make optimal transport couplings more useful for down-stream tasks, we introduce the concept of entropic filtering to retain only high-confidence correspondences. In the second step, USHER learns a low-complexity transformation that reliably restores the models representation space for OOD data. We demonstrate this learned transformation generalizes to other OOD data from similar experimental conditions. We applied USHER to correct platform-specific biases seen when running scGPT on Xenium transcript counts: USHER maps Xenium embeddings back to the native scRNA-seq representation space, improving cell type clustering and cross-platform integration. Histopathology foundation models trained on H&E images fail on MALDI metabolite-profiled tissue images due to data-acquisition artifacts. USHER corrects these, enabling cell-type classification and protein abundance imputation. USHER offers a generalizable framework to make biological foundation models portable across a rapidly-evolving experimental landscape.

bioinformatics↗
bioRxiv2025

Diffusion-based Representation Integration for Foundation Models Improves Spatial Transcriptomics Analysis

MotivationWe propose DRIFT, a framework that integrates spatial context into the input representations for foundation models by leveraging diffusion on spatial graphs derived from spatial transcriptomics (ST) data. ST captures gene expression profiles while preserving spatial context, enabling downstream analysis tasks such as cell-type annotation, clustering, and cross-sample alignment. However, due to its emerging nature, there are very few foundation models that can utilize ST data to generate embeddings generalizable across multiple tasks. Meanwhile, well-documented foundational models trained on large-scale single-cell gene expression (scRNA-seq) data have demonstrated generalizable performance across scRNA-seq assays, tissues, and tasks; however, they do not leverage the spatial information in ST data. We use heat kernel diffusion to propagate embeddings across spatial neighborhoods, incorporating the local neighborhood context of the ST data while preserving the transcriptomic representations learned by state-of-the-art single-cell foundation models. ResultsWe systematically benchmark five foundational models (both scRNA-seq and ST-based) across key ST tasks such as annotation, alignment, and clustering, ensuring a comprehensive evaluation of our proposed framework. Our results show that DRIFT significantly improves the performance of existing foundational models on ST data over specialized state-of-the-art methods. Overall, DRIFT is an effective, accessible, and generalizable framework that bridges the gap toward universal models for modeling spatial transcriptomics. Availability and ImplementationCode and data available at https://github.com/rsinghlab/DRIFT. Contactritambhara@brown.edu Supplementary informationSupplementary notes are provided with the manuscript.

bioinformatics↗