76555 research outputs found
Sort by
The use of AI to Predict Immune Subtype from Tumor Images
No full textArtificial intelligence (AI) has emerged as a powerful tool in biomedical research, with foundational models (FMs) offering the potential to extract meaningful patterns from complex data such as histopathological images. In this study, we utilize Virchow, a large-scale FM trained on over 1.4 million hematoxylin and eosin (H&E) whole slide images (WSIs), to predict tumor immune subtypes across five cancer types using a custom attention-based multiple instance learning (att-MIL) aggregator network. Tumor immune subtypes, as defined by transcriptomic profiling in prior work by Thorsson et al., are known to correlate with response to immunotherapy but are typically expensive and time-consuming to determine using RNA-seq. Our pipeline processes WSIs from The Cancer Genome Atlas (TCGA), extracts representative image tiles, generates matrix embeddings using Virchow, and uses these embeddings to train an att-MIL. The model achieved a 47.06% overall accuracy in predicting immune subtype, substantially outperforming random guessing (20%). Accuracy by subtype varied, with the highest accuracy being subtype 2 (63.48%) and no correct predictions for the less frequent subtype 6. These results demonstrate proof of concept that FM derived image embeddings can support prediction of molecularly defined immune phenotypes from histological images alone. This approach could offer a scalable, cost effective alternative to transcriptomic profiling, potentially aiding clinical decision making in immunotherapy
Unsupervised Machine Learning for Visual Frailty Prediction in Mice
No full textFrailty quantifies biological aging and predicts adverse health outcomes, but manual frailty index (FI) assessments are labor-intensive and variable across scorers. We previously developed the visual frailty index (vFI), which uses machine vision to estimate FI from video-based behavioral features. Here, we developed the Unsupervised Visual Frailty Index (uvFI), which uses unsupervised behavioral segmentation of open-field videos from 851 mice (540 C57BL/6J, 311 Diversity Outbred; 1126 trials). We used an unsupervised behavior segmentation model to extracted data-driven features from behavioral syllables, transitions, and poses to predict frailty (uvFI), age (uvFRIGHT), and proportion of life lived (uvPLL). We achieved a mean absolute error (MAE) of 1.37 ± 0.122 for frailty and 14.9 ± 1.18 weeks for age, with accuracy further improved by combining supervised and unsupervised features. These results show that unsupervised behavioral features capture aging signatures, enabling scalable and reproducible frailty assessment
A corpus of GA4GH phenopackets: Case-level phenotyping for genomic diagnostics and discovery.
No full textThe Global Alliance for Genomics and Health (GA4GH) Phenopacket Schema was released in 2022 and approved by ISO as a standard for sharing clinical and genomic information about an individual, including phenotypic descriptions, numerical measurements, genetic information, diagnoses, and treatments. A phenopacket can be used as an input file for software that supports phenotype-driven genomic diagnostics and for algorithms that facilitate patient classification and stratification for identifying new diseases and treatments. There has been a great need for a collection of phenopackets to test software pipelines and algorithms. Here, we present Phenopacket Store. Phenopacket Store v.0.1.19 includes 6,668 phenopackets representing 475 Mendelian and chromosomal diseases associated with 423 genes and 3,834 unique pathogenic alleles curated from 959 different publications. This represents the first large-scale collection of case-level, standardized phenotypic information derived from case reports in the literature with detailed descriptions of the clinical data and will be useful for many purposes, including the development and testing of software for prioritizing genes and diseases in diagnostic genomics, machine learning analysis of clinical phenotype data, patient stratification, and genotype-phenotype correlations. This corpus also provides best-practice examples for curating literature-derived data using the GA4GH Phenopacket Schema
Non-neutralizing antibodies to influenza A matrix-protein-2-ectodomain are broadly effective therapeutics and resistant to viral escape mutations.
No full textInfluenza A viruses remain a global health threat, yet no universal antibody therapy exists. Clinical programs have centered on neutralizing mAbs, only to be thwarted by strain specificity and rapid viral escape. We instead engineered three non-neutralizing IgG2a mAbs that target distinct, overlapping epitopes within the conserved N terminus of the M2 ectodomain (M2e). Combined at low dose, this triple M2e-mAb confers robust prophylactic and therapeutic protection in mice challenged with diverse human and zoonotic IAV strains, including highly pathogenic variants. Therapeutic efficacy depends on Fc-mediated effector activity via FcγRI, FcγRIII, and FcγRIV, rather than in vitro neutralization. Serial passaging in triple M2e-mAb-treated immunocompetent and immunodeficient hosts failed to generate viral escape mutants. Our findings redefine the influenza-specific antibody therapeutic design and support Fc-optimized, non-neutralizing M2e-mAbs as a broadly effective, mutation-resistant, off-the-shelve therapy with direct relevance to human pandemic preparedness
Evidence for coopetition at the maternal-fetal interface shaping placental invasion.
No full textCoopetition is a term from game theory that describes a mix of cooperative and competitive behavior. The maternal-fetal interface (MFI) among eutherian mammals presents close interaction of two distinct individuals. These interactions have resulted in a remarkable diversity in MFI structure, often interpreted as the outcome of maternal-fetal conflict. Nevertheless, the fetus and the mother share evolutionary interests since 50% of fetal genes are maternal. In hemochorial species, characterized by invasive placentation, endometrial stromal fibroblasts (ESFs) undergo decidualization to regulate embryo implantation. In great apes, hemochorial placentation is driven by highly invasive extravillous trophoblast cells (EVT). Here, using EVTs differentiated from trophoblast stem cells, term placenta, as well as HTR8/SVneo, we demonstrate that EVTs orchestrate a transformation of the maternal stroma, reducing its resistance to invasion acquired during decidualization. Through paracrine signals, in particular IL-11, trophoblasts transform decidual ESFs from a matrix-producing to a matrix-degrading state. Notably, noninvasive cytotrophoblast cells, do not transform decidual ESFs. We further provide evidence that maternal coadaptation is critical to EVT-induced decidual transformation. Decidual ESFs upregulate expression of Suppressor of Cytokine Signaling 3 in response to EVTs, rewiring downstream IL-11 signaling from JAK/STAT to AP-1 specific transcription. We conclude that the evolution of highly invasive placentation is the outcome of both the evolution of invasive EVTs, as well as the evolution of maternal traits, i.e., the switch from JAK/STAT to AP-1 signaling. We interpret this as evidence for co-opetition (cooperation among competitors)
Inflammatory stromal and T cells mediate human bone marrow niche remodeling in clonal hematopoiesis and myelodysplasia.
No full textSomatic mutations in hematopoietic stem/progenitor cells (HSPCs) can lead to clonal hematopoiesis of indeterminate potential (CHIP) and progression to myelodysplastic syndromes (MDS). Using single-cell and anatomical profiling of a large cohort of human bone marrow (BM), we show that the HSPC BM niche in CHIP and MDS is undergoing inflammatory remodeling. This includes loss of CXCL12⁺ adipogenic stromal cells and the emergence of a distinct population of inflammatory mesenchymal stromal cells (iMSCs), which arise in CHIP and become more prevalent in MDS. Functional studies in primary BM HSPC-MSC co-cultures reveals that healthy aged and CHIP HSPCs activate stromal support, while MDS HSPCs fail to do so. In contrast, MDS blasts further suppress HSPC support and trigger inflammation, indicating disease-stage-specific stromal disruption. In parallel, we show that iMSCs retain partial support and angiogenic potential in MDS, coinciding with expanded BM vasculature. Additionally, we identify IFN-responsive T cells that preferentially interact with iMSCs, potentially reinforcing local inflammation. These findings position iMSCs as central mediators of early BM niche dysfunction and potential therapeutic targets for intercepting pre-malignant hematopoiesis
Advancing the science of genomic learning healthcare systems.
No full textINTRODUCTION: Identifying key characteristics of exemplar genomic learning healthcare systems (gLHS) and knowledge gaps that can be explored by collaboration among them is likely to accelerate the sharing of best practices and generation of evidence that informs the use of genomics in clinical care.
METHODS: Deliberations of an expert group convened by the National Human Genome Research Institute (NHGRI) supplemented by relevant literature.
RESULTS: Recent advances in genomic data standardization, automated clinical decision support, increased interoperability, and improved genomic technologies have enabled the development of several robust gLHS. They remain concentrated in major academic centers, however, and operate largely independently. Sharing their methods and tools would increase access to these innovations and advance the field. Several gLHS have expressed willingness to collaborate in a coalition designed to gather, evaluate, and disseminate best practices and development needs. Such a coalition has recently been formed under the leadership of NHGRI.
CONCLUSION: Increased collaboration, interoperability, and sharing of genomic information and strategies across gLHS can help define, refine, and disseminate best practices. Such cooperation can improve genomic variant curation and interpretation, diagnostic accuracy, evidence generation, and ultimately patient care through seamless integration of research as an integral component of good clinical care
Inborn errors of immunity: Manifestation, treatment, and outcome-an ESID registry 1994-2024 report on 30,628 patients.
No full textThe European Society for Immunodeficiencies patient registry (ESID-R), established in 1994, is one of the world\u27s largest databases on inborn errors of immunity (IEI). IEI are genetic disorders predisposing patients to infections, autoimmunity, inflammation, allergies, and malignancies. Treatments include antimicrobial therapy, immunoglobulin replacement, immune modulation, stem cell transplantation, and gene therapy. Data from 194 centers in 33 countries capture clinical manifestations and treatments from birth onward, with annually expected updates. This report reviews the ESID-R\u27s structure, data content, and impact. The registry includes 30,628 patient datasets (aged 0-97.9 years; median follow-up: 7.2 years; total 825,568.2 patient-years), with 13,550 cases in 15 sub-studies. It has produced 84 peer-reviewed publications (mean citation rate: 95). Findings include real-world observations of IEI diagnoses, genetic causes, clinical manifestations, treatments, and survival trends. The ESID-R fosters global collaboration, advancing IEI research and patient care. This report highlights the key role of the multinational ESID-R, led by an independent medical society, in evidence-based discovery
Neurological, metabolic and inflammatory phenotypes in a mouse model of ECHS1 deficiency.
No full textECHS1 deficiency (ECHS1D) is a rare and devastating neurometabolic disease that currently has no defined treatments. This disorder results from missense loss-of-function mutations in the ECHS1 gene that results in severe developmental delays, encephalopathy, hypotonia and early death. ECHS1 enzymatic activity is necessary for the beta-oxidation of fatty acids and the oxidation of branched-chain amino acids within the inner mitochondrial matrix. The pathogenesis of disease remains poorly understood. To expand our knowledge on disease mechanisms, we generated a novel mouse model of ECHS1D that possesses a disease-associated variant knocked-in (KI) the Echs1 allele and a knock-out (KO) of the other Echs1 allele. Neurological and metabolic abnormalities were assessed under basal conditions, and acute inflammation was tested as a potential disease driver. Mice containing KI/KI or KI/KO alleles were viable with normal development and survival, and the combined KI and KO alleles resulted in more than a 95% reduction of Echs1 protein levels. ECHS1D mice had significantly increased epileptiform EEG activity and were sensitive to seizure induction, which resulted in the death of 60% of ECHS1D mice. Power spectral analysis revealed ECHS1D mice had increased slow-wave EEG power that was associated with sleep dysfunction. Under basal conditions, energy status and mitochondrial function within the brain was unaffected, while aromatic amino acid content was increased. Markers of neuroinflammation were increased in ECHS1D mice in an age-dependent manner and acute inflammatory challenge resulted in failure to thrive and early lethality in ECHS1D mice. In conclusion, we developed a novel model of ECHS1D that can be used to study disease mechanisms and for therapeutic development
An international perspective on the future of systemic sclerosis research.
No full textSystemic sclerosis (SSc) remains a challenging and enigmatic systemic autoimmune disease, owing to its complex pathogenesis, clinical and molecular heterogeneity, and the lack of effective disease-modifying treatments. Despite a century of research in SSc, the interconnections among microvascular dysfunction, autoimmune phenomena and tissue fibrosis in SSc remain unclear. The absence of validated biomarkers and reliable animal models complicates diagnosis and treatment, contributing to high morbidity and mortality. Advances in the past 5 years, such as single-cell RNA sequencing, next-generation sequencing, spatial biology, transcriptomics, genomics, proteomics, metabolomics, microbiome profiling and artificial intelligence, offer new avenues for identifying the early pathogenetic events that, once treated, could change the clinical history of SSc. Collaborative global efforts to integrate these approaches are crucial to developing a comprehensive, mechanistic understanding and enabling personalized therapies. Challenges include disease classification, clinical heterogeneity and the establishment of robust biomarkers for disease activity and progression. Innovative clinical trial designs and patient-centred approaches are essential for developing effective treatments. Emerging therapies, including cell-based and fibroblast-targeting treatments, show promise. Global cooperation, standardized protocols and interdisciplinary research are vital for advancing SSc research and improving patient outcomes. The integration of advanced research techniques holds the potential for important breakthroughs in the diagnosis, treatment and care of individuals with SSc