Biology subjects

Ntranos, V.

Publications and source records attributed to Ntranos, V..

15 recordsLinked to original sources

bioRxiv2026

PepCL: A replay-based continual learning framework for updating peptide-MHC models

Understanding peptide-major histocompatibility complex (MHC) class I binding is critical for effective vaccine and immunotherapy design but is a combinatorially complex challenge for which prediction models have become essential. MHC ligands are typically identified at scale via untargeted mass spectrometry (MS), and this has built a strong base for peptide-MHC model training. However, MS incompletely captures the vast peptide-MHC space due to technical, sampling, and biological biases. Although recently developed experimental assays have queried such blind spots yielding complementary information, existing peptide-MHC predictors have not yet incorporated these orthogonal data and are not designed to be updated as new data are generated. Here, we introduce PepCL (Peptide-MHC Continual Learning), a continual learning framework for updating peptide-MHC predictors with new assay data while explicitly preserving prior MS knowledge. To enable PepCL, we also develop MHCPrime, a new state-of-the-art pan-allelic peptide-MHC prediction model, trained on publicly available MS data, that can be effectively updated under our framework. We demonstrate that PepCL allows MHCPrime to learn previously unseen, assay-specific information while preventing catastrophic forgetting that is typically observed with conventional fine-tuning. We evaluate PepCL and MHCPrime in a variety of biological contexts, including infectious disease and cancer, and show improved peptide-MHC prediction that transfers across alleles for broader applicability in clinical settings. Overall, our results establish PepCL as a flexible framework for extending the utility of peptide-MHC models by improving their predictive performance as immunopeptidomics assays continue to evolve and new data become available.

bioinformatics↗

bioRxiv2026

ApoE-Dependent Lipid Handling by Median Eminence Microglia Preserves Myelin Integrity and Metabolic Function

Microglia regulate hypothalamic control of systemic metabolism, but the mechanisms underlying their contribution remain unclear. Here, we identify a distinct apolipoprotein E (ApoE) microglial population enriched in the median eminence (ME), a brain region involved in sensing peripheral cues and metabolic regulation. These microglia engage multiple functional programs related to lipid handling, interferon signaling, and stress responses that are differentially regulated within the ME. Consumption of a Western diet (WD) increased interferon signaling and lipid accumulation in ME microglia. Expression of the human APOE4 isoform in mice exacerbated microglial lipid dysregulation, interferon signaling, and impaired ME myelin organization. Deleting APOE in microglia attenuated their ability to couple lipid accumulation to interferon signaling, identifying microglial APOE as a cell-intrinsic determinant of interferon responses. Finally, selective activation of liver X receptor signaling using synthetic HDL nanoparticles restored microglial lipid homeostasis, improved hypothalamic leptin responsiveness, and limited weight gain in WD-fed mice. Together, these findings define an Apoe-dependent regulatory program in ME microglia that is therapeutically targetable and clarify how nutritional stress disrupts hypothalamic control of metabolic homeostasis.

neuroscience↗

bioRxiv2026

Fezf2 regulates differentiation of Aire-expressing and post-Aire mimetic epithelial populations maintaining thymic homeostasis

Fezf2 has been proposed to serve principally as a transcriptional regulator of broad self-antigen expression in medullary thymic epithelial cells (mTECs). Here, we demonstrate an additional function for Fezf2 as a developmental regulator of TEC lineage commitment. Fezf2 deficiency in the thymus led to a relative expansion of lung mimetic and Ccl21+ immature mTECs at the expense of MHCII-hi Aire-expressing and all other mimetic populations. Consistent with this, the disruption of mTEC development had functional consequences, including alterations in tuft-associated iNKT polarization and microfold-associated B-cell class switching. In addition, high-resolution transcriptomics revealed that Fezf2 and Aire regulate distinct transcriptional programs, with Fezf2 driving both activation and repression of a more limited set of tissue-restricted genes compared to the broad gene activation program observed for Aire. Pure transcriptional repression at Fezf2 target loci was sufficient to partially rescue mTEC development in global Fezf2-/- (tKO) mice, potentially through suppression of Lifr expression and modulation of downstream Stat3 signaling tone. These results expand our understanding of Fezf2 in mTEC biology, highlighting transcriptional repression as a required functional facet for adult steady-state lineage patterning across the mTEC compartment.

immunology↗

bioRxiv2025

Blocking Osteoprotegerin Reprograms Cancer Associated Fibroblast to Promotes Immune Infiltration into the Tumor Microenvironment

AbstractThe stromal compartment of many solid tumors plays a critical role in shaping an immunosuppressive microenvironment that limits the effectiveness of immune-based therapies1. Among stromal constituents, cancer-associated fibroblasts (CAFs) have emerged as key regulators of antitumor immunity2-5. Here, we identify a distinct subset of CAFs in both murine and human stroma-rich cancers that secrete osteoprotegerin (OPG)-- a soluble decoy receptor that neutralizes receptor activator of nuclear factor kappa-B ligand (RANKL) and TNF-related apoptosis-inducing ligand (TRAIL), both of which are involved in T cell function. In vitro, OPG directly impairs CD8 T cell-mediated killing of target cells. In murine models of pancreatic and breast cancer, antibody-mediated blockade of OPG promotes robust immune infiltration into the tumor microenvironment, leading to significant tumor regression. Stromal profiling revealed that OPG blockade induces a shift in CAF cells--reducing immunosuppressive OPG fibroblasts while expanding interferon-responsive fibroblasts, thus recalibrating the tumor stroma toward a pro-immunogenic landscape. These findings uncover a previously unrecognized mechanism of stromal immune suppression and highlight OPG as a stromal immune checkpoint controlling CD8 T cell infiltration. Targeting OPG may offer a novel therapeutic strategy to convert immunologically "cold" tumors into T cell-infiltrated, tumor microenvironment.

cancer biology↗

bioRxiv2025

Phenotypic pleiotropy of missense variants in human B cell-confinement receptor P2RY8

Missense variants can have pleiotropic effects on protein function and predicting these effects can be difficult. We performed near-saturation deep mutational scanning of P2RY8, a G-protein-coupled receptor that promotes germinal center B cell confinement. We assayed the effect of each variant on surface expression, migration, and proliferation. We delineated variants that affected both expression and function, affected function independently of expression, and discrepantly affected migration and proliferation. We also used cryo-electron microscopy to determine the structure of activated, ligand-bound P2RY8, providing structural insights into the effects of variants on ligand binding and signal transmission. We applied the deep mutational scanning results to both improve computational variant effect predictions and to characterize the phenotype of germline variants and lymphoma-associated variants. Together, our results demonstrate the power of integrating deep mutational scanning, structure determination, and in silico prediction to advance the understanding of a receptor important in human health.

genomics↗

bioRxiv2024

nSMase2-mediated exosome secretion shapes the tumor microenvironment to immunologically support pancreatic cancer

The pleiotropic roles of nSMase2-generated ceramide include regulation of intracellular ceramide signaling and exosome biogenesis. We investigated the effects of eliminating nSMase2 on early and advanced PDA, including its influence on the microenvironment. Employing the KPC mouse model of pancreatic cancer, we demonstrate that pancreatic epithelial nSMase2 ablation reduces neoplasia and promotes a PDA subtype switch from aggressive basal-like to classical. nSMase2 elimination prolongs survival of KPC mice, hinders vasculature development, and fosters a robust immune response. nSMase2 loss leads to recruitment of cytotoxic T cells, N1-like neutrophils, and abundant infiltration of anti-tumorigenic macrophages in the pancreatic preneoplastic microenvironment. Mechanistically, we demonstrate that nSMase2-expressing PDA cell small extracellular vesicles (sEVs) reduce survival of KPC mice; PDA cell sEVs generated independently of nSMase2 prolong survival of KPC mice and reprogram macrophages to a proinflammatory phenotype. Collectively, our study highlights previously unappreciated opposing roles for exosomes, based on biogenesis pathway, during PDA progression. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=189 SRC="FIGDIR/small/614610v1_ufig1.gif" ALT="Figure 1"> View larger version (43K): org.highwire.dtl.DTLVardef@b6e6d1org.highwire.dtl.DTLVardef@1ebb41eorg.highwire.dtl.DTLVardef@17e4275org.highwire.dtl.DTLVardef@3d31f1_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2024

Uncovering differential tolerance to deletions versus substitutions with a protein language model

Deep mutational scanning (DMS) experiments have been successfully leveraged to understand genotype to phenotype mapping, with broad implications for protein engineering, human genetics, drug development, and beyond. To date, however, the overwhelming majority of DMS have focused on amino acid substitutions, excluding other classes of variation such as deletions or insertions. As a consequence, it remains unclear how indels differentially shape the fitness landscape relative to substitutions. In order to further our understanding of the relationship between substitutions and deletions, we leveraged a protein language model to analyze every single amino acid deletion in the human proteome. We discovered hundreds of thousands of sites that display opposing behavior for deletions versus substitutions, i.e. sites that can tolerate being substituted but not deleted, and vice versa. We identified secondary structural elements and sequence context to be important mediators of differential tolerability at these sites. Our results underscore the value of deletion-substitution comparisons at the genome-wide scale, provide novel insights into how substitutions could systematically differ from deletions, and showcase the power of protein language models to generate biological hypotheses in-silico. All deletion-substitution comparisons can be explored and downloaded at https://huggingface.co/spaces/ntranoslab/diff-tol.

bioinformatics↗

bioRxiv2023

Single cell sequencing as a general variant interpretation assay

The human genome contains [~]70 million possible protein-altering variants, the vast majority of which are of uncertain clinical significance. Closing this gap is essential for accurate diagnosis of disease-causing variants and understanding their mechanisms of action. Towards this goal, we developed a pooled perturbation approach combining saturation mutagenesis with single cell RNA sequencing to map the effects of every single nucleotide variant in a gene. We sequenced [~]440,000 cells expressing variants in CDKN2A (p16INK4a), TP53, and SOD1, observing almost all possible protein-coding variants, with a mean of 61 cells per variant. Using single cell gene expression signatures, we show that each gene may contain multiple types of pathogenic variants that affect distinct downstream pathways. We demonstrate that single cell expression signatures outperform existing bulk experimental assays and computational models for predicting pathogenicity, and summarize both the utility and potential limitations of single cell sequencing as a general variant interpretation assay.

genomics↗

bioRxiv2022

memento: Generalized differential expression analysis of single-cell RNA-seq with method of moments estimation and efficient resampling

Differential expression analysis of scRNA-seq data is central for characterizing how experimental factors affect the distribution of gene expression. However, it remains challenging to distinguish biological and technical sources of cell-cell variability and to assess the statistical significance of quantitative comparisons between groups of cells. We introduce memento to address these limitations and enable accurate and efficient differential expression analysis of the mean, variability, and gene correlation from scRNA-seq. We used memento to analyze 70,000 tracheal epithelial cells to identify interferon response genes with distinct variability and correlation patterns, 160,000 T cells perturbed with CRISPR-Cas9 to reconstruct gene-regulatory networks that control T cell activation, and 1.2 million PMBCs to map cell-type-specific cis expression quantitative trait loci (eQTLs). In all cases, memento identified more significant and reproducible differences in mean expression but also identified differences in variability and gene correlation that suggest distinct modes of transcriptional regulation imparted by cytokines, genetic perturbations, and natural genetic variation. These results demonstrate memento as a first-in-class method for the quantitative comparisons of scRNA-seq data scalable to millions of cells and thousands of samples.

bioinformatics↗

bioRxiv2022

Genome-wide prediction of disease variants with a deep protein language model

Distinguishing between damaging and neutral missense variants is an ongoing challenge in human genetics, with profound implications for clinical diagnosis, genetic studies and protein engineering. Recently, deep-learning models have achieved state-of-the-art performance in classifying variants as pathogenic or benign. However, these models are currently unable to provide predictions over all missense variants, either because of dependency on close protein homologs or due to software limitations. Here we leveraged ESM1b, a 650M-parameter protein language model, to predict the functional impact of human coding variation at scale. To overcome existing technical limitations, we developed a modified ESM1b workflow and functionalized, for the first time, all proteins in the human genome, resulting in predictions for all [~]450M possible missense variant effects. ESM1b was able to distinguish between pathogenic and benign variants across [~]150K variants annotated in ClinVar and HGMD, outperforming existing state-of-the-art methods. ESM1b also exceeded the state of the art at predicting the experimental results of deep mutational scans. We further annotated [~]2M variants across [~]9K alternatively-spliced genes as damaging in certain protein isoforms while neutral in others, demonstrating the importance of considering all isoforms when functionalizing variant effects. The complete catalog of variant effect predictions is available at: https://huggingface.co/spaces/ntranoslab/esm_variants.

bioinformatics↗

bioRxiv2021

Single-Cell Multiomics Defines Tolerogenic Extrathymic Aire-Expressing Populations with Unique Homology to Thymic Epithelium

The Autoimmune Regulator (Aire) gene, well defined for its role in medullary thymic epithelial cells (mTECs) and immune self-tolerance, is also expressed in extrathymic Aire-expressing cells (eTACs) in the secondary lymphoid organs. eTACs have been shown to be hematopoietic antigen presenting cells (APCs) and potent inducers of immune tolerance (1-3). However, the precise identity and function of these cells remain unclear. Here, we use high-dimensional single-cell multiomics and functional approaches to define eTACs at the transcriptional, genomic, and proteomic level. We find that eTACs consist of two similar cell types: CCR7+ Aire-expressing migratory dendritic cells (AmDCs) and a unique Aire-hi population co-expressing Aire and RAR-related orphan receptor gamma-t (ROR{gamma}t). The latter, which have significant transcriptional and genomic homology to migratory dendritic cells (migDCs) and mTECs, we term Janus cells (JCs). All eTACs, and JCs in particular, have a highly accessible chromatin structure and high levels of broad gene expression, including tissue-specific antigens, as well as remarkable transcriptional and genomic homology to thymic medullary epithelium. As in the thymus, Aire expression in eTACs is also dependent on RANK-RANK-ligand interactions. Furthermore, lineage-tracing shows that JCs are not precursors to the majority of AmDCs. Finally, self-antigen expression by eTACs is sufficient to mediate negative selection of T cells escaping thymic selection and can prevent autoimmune diabetes in non-obese diabetic mice. This transcriptional, genomic, and functional symmetry between a hematopoietic Aire-expressing population in the periphery and an epithelial Aire-expressing population in the thymus suggests that a core biological program may influence self-tolerance and self-representation across the spectrum of immune development.

immunology↗

bioRxiv2018

The Barcode, UMI, Set format and BUStools

We introduce the Barcode-UMI-Set format (BUS) for representing pseudoalignments of reads from single-cell RNA-seq experiments. The format can be used with all single-cell RNA-seq technologies, and we show that BUS files can be efficiently generated. BUStools is a suite of tools for working with BUS files and facilitates rapid quantification and analysis of single-cell RNA-seq data. The BUS format therefore makes possible the development of modular, technology-specific, and robust workflows for single-cell RNA-seq analysis.

bioinformatics↗

bioRxiv2018

One read per cell per gene is optimal for single-cell RNA-Seq

An underlying question for virtually all single-cell RNA sequencing experiments is how to allocate the limited sequencing budget: deep sequencing of a few cells or shallow sequencing of many cells? A mathematical framework reveals that, for estimating many important gene properties, the optimal allocation is to sequence at the depth of one read per cell per gene. Interestingly, the corresponding optimal estimator is not the widely-used plugin estimator but one developed via empirical Bayes.

bioinformatics↗

bioRxiv2018

Identification of transcriptional signatures for cell types from single-cell RNA-Seq

Single-cell RNA-Seq makes it possible to characterize the transcriptomes of cell types and identify their transcriptional signatures via differential analysis. We present a fast and accurate method for discriminating cell types that takes advantage of the large numbers of cells that are assayed. When applied to transcript compatibility counts obtained via pseudoalignment, our approach provides a quantification-free analysis of 3 single-cell RNA-Seq that can identify previously undetectable marker genes.

bioinformatics↗

bioRxiv2017

Column subset selection for single-cell RNA-Seq clustering

The first step in the analysis of single-cell RNA sequencing (scRNA-Seq) is dimensionality reduction, which reduces noise and simplifies data visualization. However, techniques such as principal components analysis (PCA) fail to preserve non-negativity and sparsity structures present in the original matrices, and the coordinates of projected cells are not easily interpretable. Commonly used thresholding methods avoid those pitfalls, but ignore collinearity and covariance in the original matrix. We show that a deterministic column subset selection (DCSS) method possesses many of the favorable properties of PCA and common thresholding methods, while avoiding pitfalls from both. We derive new spectral bounds for DCSS. We apply DCSS to two measures of gene expression from two scRNA-Seq experiments with different clustering workflows, and compare to three thresholding methods. In each case study, the clusters based on the small subset of the complete gene expression profile selected by DCSS are similar to clusters produced from the full set. The resulting clusters are informative for cell type.

bioinformatics↗