Biology subjects

Schwartz, L.

Publications and source records attributed to Schwartz, L..

12 recordsLinked to original sources

bioRxiv2026

ExPert infers causal perturbations from cell states across biological and clinical contexts

Pooled single-cell CRISPR screens have driven substantial effort toward models predicting the response to a perturbation. We posed the inverse question: which perturbation produced an observed cell state? Here we developed ExPert (Extracting Perturbations), which learns one prototype per perturbation in a latent space shared by 12 cellular contexts, trained on 8,964 perturbations and 19.5 million cells, and ranks 19,895 candidate genes per query, over half absent from training. ExPert more than doubled held-out-context recall over the best existing method with 1,000-fold less compute, and transferred without retraining to drugs, protein degradation and combinations. In alcohol-associated liver disease it implicated mRNA export, unchanged in expression, and we experimentally confirmed nuclear poly(A) retention and isoform switching. In primary mitochondrial disease it recovered the affected complex from blood, localized the signal to monocytes, and flagged a misdiagnosed control. Perturbation atlases can thus be queried in reverse, from state to likely driver.

systems biology↗

bioRxiv2025

Interpretable Thermodynamic Score-based Classification of Relaxation Excursions

Classifying cellular states from high-dimensional molecular and genomic measurements requires methods that provide not only accurate predictions but also calibrated uncertainty and interpretability. Current nonlinear classifiers offer accuracy but often lack uncertainty quantification and mechanistic insights into the features that matter most. We introduce Keeping SCORE, a framework that transforms conditional diffusion models into probabilistic engines for classification and regression by computing exact likelihoods along stochastic noising trajectories. We first benchmark Keeping SCORE on image recognition tasks (handwritten digits, natural photos). We then apply Keeping SCORE to single-cell transcriptomics across a 22-million-cell atlas, classifying 164 cell types with accuracy matching or exceeding state-of-the-art methods, while uniquely providing posterior probability estimates and prediction confidence. For genetic perturbation mapping across 100 CRISPRi conditions in a multi-study Perturb-seq dataset, our approach again matches or surpasses discriminative baselines, with feature-level attributions identifying which genomic features drive each decision. Applied to large-scale protein sequence data, our framework accurately regresses mutational stability effects, attributing them quantitatively to positions along the input sequence. Keeping SCORE requires no retraining or architectural changes to existing diffusion models, providing portable, interpretable, and uncertainty-aware predictions for biological discovery.

bioinformatics↗

bioRxiv2025

Fluctuation structure predicts genome-wide perturbation outcomes

Pooled single-cell perturbation screens represent powerful experimental platforms for functional genomics, yet interpreting these rich datasets for meaningful biological conclusions remains challenging. Most current methods fall at one of two extremes: either opaque deep learning models that obscure biological meaning, or simplified frameworks that treat genes as isolated units. As such, these approaches overlook a crucial insight: gene co-fluctuations in unperturbed cellular states can be harnessed to model perturbation responses. Here we present CIPHER (Covariance Inference for Perturbation and High-dimensional Expression Response), a conceptual framework leveraging ideas from linear response in statistical physics to model transcriptome-wide perturbation outcomes using gene co-fluctuations in unperturbed cells. We validated our approach on synthetic regulatory networks before applying it to 29 large-scale single-cell genetic perturbation datasets covering 19,003 perturbations and over 6.26M cells. Our work robustly recapitulated genome-wide responses to single and double perturbations by exploiting baseline gene covariance structure. Importantly, eliminating gene-gene covariances, while retaining gene-intrinsic variances, i.e. mean-field conditions, dramatically reduced model performance by several folds across multiple metrics, demonstrating the rich information stored within baseline fluctuation structures. Benchmarked against recent deep learning and linear baselines, CIPHER matched or exceeded the best-performing approaches with fitting a single parameter. Moreover, gene-gene correlations transferred successfully across independent studies of the same cell type, revealing stereotypic fluctuation structures. We further extended CIPHER to the inverse problem of identifying true driver perturbations, where it achieved high performance across both genetic and chemical perturbation screens through uncertainty-aware Bayesian inference. We further used CIPHER to nominate drivers of therapy resistance in melanoma and pancreatic cancer, validating its top predictions experimentally. Finally, most genome-wide responses propagated through the covariance matrix along approximately 1-3 independent and global gene modules, consistent with a low-rank structure of the underlying gene regulatory network, which we show can enable the framework's success. We have also created a package called cipher-perturb, available on PyPI, to apply the framework to any dataset, accompanied by a detailed website (https://goyallab.github.io/CIPHERWebsite/). Our study underscores the importance of theoretically-grounded models in capturing complex biological responses, highlighting fundamental design principles encoded in cellular fluctuation patterns.

systems biology↗

bioRxiv2025

Obesity promotes urinary tract infection by disrupting urothelial immune defenses

Obesity is a significant public health concern that is associated with numerous health risks. Infections are a major complication of obesity, but the mechanisms responsible for increased infection risk are poorly defined. Here, we use a diet induced obesity mouse model and investigate how obesity impacts urinary tract infection (UTI) susceptibility and bladder immune defenses. Our results show that high-fat diet fed female and male mice exhibit increased susceptibility to uropathogenic E. coli (UPEC) following experimental UTI. Transcriptomic analysis of bladder urothelial cells shows that obesity alters gene expression in a sex-specific manner, with distinct differentially expressed genes in male and female mice, but shared activation of focal adhesion and extracellular matrix signaling. Western blot and immunostaining confirm activation of focal adhesion kinase, a central component of the focal adhesion pathway, in the bladders of obese female and male mice. Mechanistically, experiments using primary human urothelial cells demonstrate that focal adhesion kinase overexpression promotes UPEC invasion. These findings demonstrate that obesity enhances UTI susceptibility by activating focal adhesion kinase and promoting bacterial invasion of the urothelium. Together, they explain how obesity promotes UTI vulnerability and identify modifiable targets for managing obesity-associated UTI.

cell biology↗

bioRxiv2024

Cancer Interception During Treatment: Using Growth Kinetics to Create a Continuous Variable for Assessing Disease Response

BackgroundWe applied 11 mathematical models of tumor growth to clinical trial data available from public and private sources. We have previously described the remarkable capacity for a simple biexponential model of tumor growth to fit thousands of datasets, and to correlate with overall survival. The goal of this study was to extend our analysis to additional tumor types and to evaluate whether alternate growth models could describe the time course of disease burden in the small subset of patients in whom the biexponential model failed. MethodsFor this analysis, we obtained data for tumor burden from 17,140 patients with six different tumor types. Imaging data and serum levels of tumor markers were available for 3,346 and 13,794 patients, respectively. Data from patients were first analyzed using the biexponential model to determine rates of tumor growth (g) and regression (d); for those whose data could not be described by this model, fit of their data was assessed using seven alternative models. The model that minimized the Akaike Information Criterion was selected as the best fit. Using the model that best fit an individual patients data, we estimated the rates of growth (g) and regression (d) of disease burden over time. The rates of tumor growth (g) were examined for association with a traditional endpoint (overall survival). FindingsFor each model, the number of patient datasets that fit the model were obtained. As we have previously reported, data from most patients fit a simple model of exponential growth and exponential regression (86%). Data from another 7% of patients fit an alternative model, including 3% fitting to a model of constant or linear regression and exponential growth of tumor on the surface and 3% fitting to model of exponential decay on tumor surface with asymmetric growth. As previously reported, we found that growth rate correlates well with overall survival, remarkably even when data from various histologies are considered together. For patients with multiple timepoints of tumor measurement, the growth rate could often be estimated even during the phase when only net regression of tumor quantity could be discerned. InterpretationThe validation of a simple mathematical model across different cancers and its application to existing clinical data allowed estimation of the rate of growth of a treatment resistant subpopulation of cancer cells. The quantification of available clinical data using the growth rate of tumors in individual patients and its strong correlation with overall survival makes the growth rate an excellent marker of the efficacy of therapy specific to the individual patient.

cancer biology↗

bioRxiv2024

Broad de-regulated U2AF1 splicing is prognostic and augments leukemic transformation via protein arginine methyltransferase activation

The role of splicing dysregulation in cancer is underscored by splicing factor mutations; however, its impact in the absence of such rare mutations is poorly understood. To reveal complex patient subtypes and putative regulators of pathogenic splicing in Acute Myeloid Leukemia (AML), we developed a new approach called OncoSplice. Among diverse new subtypes, OncoSplice identified a biphasic poor prognosis signature that partially phenocopies U2AF1-mutant splicing, impacting thousands of genes in over 40% of adult and pediatric AML cases. U2AF1-like splicing co-opted a healthy circadian splicing program, was stable over time and induced a leukemia stem cell (LSC) program. Pharmacological inhibition of the implicated U2AF1-like splicing regulator, PRMT5, rescued leukemia mis-splicing and inhibited leukemic cell growth. Genetic deletion of IRAK4, a common target of U2AF1-like and PRMT5 treated cells, blocked leukemia development in xenograft models and induced differentiation. These analyses reveal a new prognostic alternative-splicing mechanism in malignancy, independent of splicing-factor mutations. Statement of significanceUsing a new in silico strategy we reveal counteracting determinants of patient survival in Acute Myeloid Leukemia that co-opt well-defined mutation-dependent splicing programs. Broad poor-prognosis splicing and leukemia stem cell survival could be rescued through pharmacological inhibition (PRMT5) or target deletion (IRAK4), opening the door for new precision therapies. Competing InterestsConflict-of-interest disclosure: DTS. serves on the scientific advisory board at Kurome Therapeutics; is a consultant for and/or received funding from Kurome Therapeutics, Captor Therapeutics, Treeline Biosciences, and Tolero Therapeutics; and has equity in Kurome Therapeutics. AV has received research funding from GlaxoSmithKline, BMS, Jannsen, Incyte, MedPacto, Celgene, Novartis, Curis, Prelude and Eli Lilly and Company, has received compensation as a scientific advisor to Novartis, Stelexis Therapeutics, Acceleron Pharma, and Celgene, and has equity ownership in Throws Exception and Stelexis Therapeutics.

cancer biology↗

bioRxiv2023

Characterization of an Osmr Conditional Knockout Mouse Model

Oncostatin M (OSM) is a member of the interleukin-6 (IL-6) family of cytokines and has been found to have distinct anti-inflammatory and pro-inflammatory properties in various cellular and disease contexts. OSM signals through two receptor complexes, one of which includes OSMR{beta}. To investigate OSM-OSMR{beta} signaling in adult hematopoiesis, we utilized the readily available conditional Osmrfl/fl mouse model B6;129-Osmrtm1.1Nat/J, which is poorly characterized in the literature. This model contains loxP sites flanking exon 2 of the Osmr gene. We crossed Osmrfl/fl mice to interferon-inducible Mx1-Cre, which is robustly induced in adult hematopoietic cells. We observed complete recombination of the Osmrfl allele and loss of exon 2 in hematopoietic (bone marrow) as well as non-hematopoietic (liver, lung, kidney) tissues. Using a TaqMan assay with probes downstream of exon 2, Osmr transcript was lower in the kidney but equivalent in bone marrow, lung, and liver from Osmrfl/fl Mx1-Cre versus Mx1-Cre control mice, suggesting that transcript is being produced despite loss of this exon. Western blots show that liver cells from Osmrfl/fl Mx1-Cre mice had complete loss of OSMR protein, while bone marrow, kidney, and lung cells had reduced OSMR protein at varying levels. RNA-seq analysis of a subpopulation of bone marrow cells (hematopoietic stem cells) finds that some OSM-stimulated genes, but not all, are suppressed in Osmrfl/fl Mx1-Cre cells. Together, our data suggest that the B6;129-Osmrtm1.1Nat/J model should be utilized with caution as loss of Osmr exon 2 has variable and tissue-dependent impact on mRNA and protein expression.

genetics↗

bioRxiv2023

Diurnal rhythmicity of fecal microbiota and metabolite profiles in the first year of life: a randomized controlled interventional trial with infant formula

Microbiota assembly in the infant gut is influenced by time and duration of dietary exposure to breast-milk, infant formula and solid foods. In this randomized controlled intervention study, longitudinal sampling of infant stool (n=998) showed similar development of fecal bacterial communities (16S rRNA/shallow metagenomics sequencing) between formula- and breast-fed infants during the first year of life (N=210). Infant formula supplemented with galacto-oligosaccharides (GOS) was most efficient to sustain high levels of bifidobacteria compared to formula containing B. longum and B. breve or placebo. In addition to primary endpoints, metabolite and bacterial profiling revealed 24-hour oscillations and data integration identified circadian networks. Rhythmicity in bacterial diversity, specific taxa and functional pathways increased with age and was strongest following breast-feeding and GOS-supplementation. Circadian rhythms in dominant taxa were discovered ex-vivo in a chemostat model. Hence, microbiota rhythmicity develops early in life, likely via bacterial intrinsic clock mechanism and is affected by diet.

microbiology↗

bioRxiv2023

circRNA-sponging: a pipeline for extensive analysis of circRNA expression and their role in miRNA sponging

MotivationCircular RNAs (circRNAs) are long non-coding RNAs (lncRNAs) often associated with diseases and considered potential biomarkers for diagnosis and treatment. Among other functions, circRNAs have been shown to act as microRNA (miRNA) sponges, preventing the role of miRNAs that repress their targets. However, there is no pipeline to systematically assess the sponging potential of circRNAs. ResultsWe developed circRNA-sponging, a nextflow pipeline that (1) identifies circRNAs via backsplicing junctions detected in RNA-seq data, (2) quantifies their expression values in relation to their linear counterparts spliced from the same gene, (3) performs differential expression analysis, (4) identifies and quantifies miRNA expression from miRNA-sequencing (miRNA-seq) data, (5) predicts miRNA binding sites on circRNAs, (6) systematically investigates potential circRNA-miRNA sponging events, (7) creates a network of competing endogenous RNAs, and (8) identifies potential circRNA biomarkers. We showed the functionality of the circRNA-sponging pipeline using RNA sequencing data from brain tissues, where we identified two distinct types of circRNAs characterized by a specific ratio of the number of the binding site to the length of the transcript. The circRNA-sponging pipeline is the first end-to-end pipeline to identify circRNAs and their sponging systematically with raw total RNA-seq and miRNA-seq files, allowing us to better indicate the functional impact of circRNAs as a routine aspect in transcriptomic research. Availabilityhttps://github.com/biomedbigdata/circRNA-sponging Contactmarkus.daniel.hoffmann@tum.de; markus.list@tum.de Supplementary MaterialSupplementary data are available at Bioinformatic Advances online.

bioinformatics↗

bioRxiv2022

Technologically-Assisted Communication Attenuates Inter-Brain Synchrony

The transition to technologically-assisted communication has permeated all facets of human social life; yet, its impact on the social brain is still unknown and the effects may be most notable during key developmental transitions. Applying a two-brain perspective, the current pre-registered study measured mother-child brain-to-brain synchrony using hyperscanning EEG at the transition to adolescence during live face-to-face interaction versus technologically-assisted remote communication. The live interaction elicited 9 significant cross-brain links between densely inter-connected frontal and temporal areas in the beta range [14-30 Hz]. Mothers right frontal region connected with childs right and left frontal, temporal, and central regions, suggesting its regulatory role in organizing the two-brain dynamics. In contrast, the remote interaction elicited only 1 significant cross-brain-cross-hemisphere link, attenuating the robust right-to-right-brain connectivity during live social moments that communicates socio-affective signals. Furthermore, while the level of social behavior was comparable between the two interactions, brain-behavior links emerged only during the live exchange, suggesting that remote interactions yield a somewhat thinner biobehavioral experience. Mother-child right temporal-temporal synchrony linked with moments of shared gaze and the degree of child engagement and empathic behavior was associated with right frontal-frontal synchrony. Our findings indicate that human co-presence is underpinned by specific neurobiological processes, suggest potential reasons for "zoom fatigue", and open a much-needed discussion on the cost of social technology for brain maturation, particularly among youth. HighlightsO_LITechnologically-assisted communication is prevalent; yet, its impact on the social brain is unknown C_LIO_LIWe measured mother-child brain-to-brain synchrony during live and technologically-assisted remote interaction C_LIO_LIThe live interaction elicited 9 cross-brain links between densely inter-connected frontal and temporal areas in the beta range C_LIO_LIThe remote interaction yielded only 1 significant cross-brain cross-hemisphere link C_LIO_LIBrain-behavior linked emerged only during the live interaction C_LIO_LIFurther research should examine the cost of social technology to brain maturation, particularly among youth. C_LI

neuroscience↗

bioRxiv2022

Activation of targetable inflammatory immune signaling is seen in Myelodysplastic Syndromes with SF3B1 mutations

BackgroundMutations in the SF3B1 splicing factor are commonly seen in Myelodysplastic syndromes (MDS) and Acute Myeloid Leukemia (AML), yet the specific oncogenic pathways activated by missplicing have not been fully elucidated. Inflammatory immune pathways have been shown to play roles in pathogenesis of MDS, though the exact mechanisms of their activation in splicing mutant cases are not well understood. MethodsRNA-seq data from SF3B1 mutant samples was analyzed and functional roles of IRAK4 isoforms were determined. Efficacy of IRAK4 inhibition was evaluated in pre-clinical models of MDS/AML ResultsRNA-seq splicing analysis of innate immune mediators in SF3B1 mutant MDS samples revealed retention of full-length exon 6 of interleukin-1 receptor-associated kinase 4 (IRAK4), a critical downstream mediator that links the Myddosome to inflammatory NF-kB activation. Exon 6 retention leads to a longer isoform, encoding a protein (IRAK4-Long) that contains the entire death domain and kinase domain, leading to maximal activation of NF-kB. Cells with wild-type SF3B1 contain smaller IRAK4 isoforms that are targeted for proteosomal degradation. Expression of IRAK4-Long in SF3B1 mutant cells induces TRAF6 activation leading to K63-linked ubiquitination of CDK2, associated with a block in hematopoietic differentiation. Inhibition of IRAK4 with CA-4948, leads to reduction in NF-kB activation, inflammatory cytokine production, enhanced myeloid differentiation in vitro and reduced leukemic growth in xenograft models. ConclusionsSF3B1 mutation leads to expression of a therapeutically targetable, longer, oncogenic IRAK4 isoform in AML/MDS models.

cancer biology↗

bioRxiv2020

Heritable Genetic Background Alters Survival and Phenotype of Mll-AF9-Induced Leukemias

The MLL-AF9 fusion protein occurring as a result of t(9;11) translocation gives rise to pediatric and adult acute leukemias of distinct lineages, including acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), and mixed phenotype acute leukemia (MPAL). The mechanisms underlying how this same fusion protein results in diverse leukemia phenotypes among different individuals is not well understood. Given emerging evidence from genome-wide association studies (GWAS) that genetic risk factors contribute to MLL-rearranged leukemogenesis, here we tested the impact of genetic background on survival and phenotype of a well-characterized Mll-AF9 knockin mouse model. We crossed this model to five distinct inbred strains (129, A/J, C57BL/6, NOD, CAST), and tested their F1 hybrid progeny for dominant genetic effects on Mll-AF9 phenotypes. We discovered that genetic background altered peripheral blood composition, with Mll-AF9 CAST F1 demonstrating significantly increased B lymphocyte frequency while the remainder of the strains exhibited myeloid-biased hematopoiesis, similar to the parental line. Genetic background also impacted overall survival, with Mll-AF9 A/J F1 and Mll-AF9 129 F1 having significantly shorter survival, and Mll-AF9 CAST F1 having longer survival, compared to the parental line. Furthermore, we observed a range of hematologic malignancies, with Mll-AF9 A/J F1, Mll-AF9 129 F1 and Mll-AF9 B6 F1 developing exclusively myeloid cell malignancies (myeloproliferative disorder (MPD) and AML) whereas a subset of Mll-AF9 NOD F1 developed MPAL and Mll-AF9 CAST F1 developed ALL. This study provides a novel in vivo experimental model to evaluate the underlying mechanisms by which MLL-AF9 results in diverse leukemia phenotypes and provides definitive experimental evidence that genetic risk factors contribute to survival and phenotype of MLL-rearranged leukemogenesis.

cancer biology↗