2026-10-12T18:13:07Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34009815 2026-09-28T05:46:29Z category_12 portal_316 item_type_3 month_year_09_2026

Data Sheet 1_Integrated pathogen detection and immune profiling in posner-schlossman syndrome: a metagenomic and cytokine approach.csv Dongqing Yuan (5659165) Chenfeng Gu (25120128) Yuan Fang (132924) Qin Yang (25816) Cell Biology aqueous humor cytokine profiling cytomegalovirus inflammation amplifier machine learning metagenomic next-generation sequencing posner-schlossman syndrome Background<p>Posner-Schlossman syndrome (PSS) is recurrent unilateral anterior uveitis with severely elevated intraocular pressure disproportionate to visible inflammation. Cytomegalovirus (CMV) is linked to partial PSS cases, yet it is unknown whether CMV induces unique immune signatures or only amplifies baseline inflammation. We combined metagenomic next-generation sequencing (mNGS), multiplex cytokine detection and machine learning to clarify CMV’s pathogenic role.</p>Methods<p>This single-center prospective study recruited 52 untreated patients with acute PSS and 25 sex-matched cataract controls. Aqueous humor samples underwent mNGS pathogen testing to divide PSS patients into CMV-positive and CMV-negative groups, and cytometric bead array was used to quantify 9 immune mediators. Seven machine learning algorithms covering linear, kernel, ensemble tree, probabilistic and discriminant models were applied to distinguish CMV status, with XGBoost and LightGBM added to validate result stability. All models were built in Python 3.9.</p>Results<p>mNGS identified CMV in 42.3% of PSS patients. Baseline clinical features were comparable between CMV subgroups. Six pro-inflammatory cytokines were significantly higher in PSS patients than controls, whereas no cytokine differed between CMV subgroups after Bonferroni correction. All seven models poorly discriminated CMV status; quadratic discriminant analysis yielded the maximum AUC of merely 0.552. SHAP values verified weak predictive power of cytokines, and learning curves indicated severe overfitting and limited generalizability within the evaluated range of training sample sizes.</p>Conclusion<p>Within the current sample size and measured cytokine panel, no statistically significant or stable difference in aqueous cytokine profiles was detected between CMV-positive and CMV-negative PSS patients, even with advanced machine learning methods. Our findings are consistent with the inflammation amplifier hypothesis, but the current cross-sectional data cannot definitively confirm CMV’s biological role, nor rule out CMV as a primary pathogenic trigger in a subset of PSS cases. Direct pathogen detection is required for personalized antiviral treatment, and shared inflammatory pathways provide universal targets for novel immunomodulatory therapies for all PSS patients.</p> 2026-09-28T05:46:29Z Dataset Dataset 10.3389/fcell.2026.1918795 https://figshare.com/articles/dataset/Data_Sheet_1_Integrated_pathogen_detection_and_immune_profiling_in_posner-schlossman_syndrome_a_metagenomic_and_cytokine_approach_csv/34009815 CC BY 4.0