Search the repository

A+ | A- | | SLO | ENG

Abstract: Marine prokaryotic communities are major contributors to oceanic food webs and global biogeochemical cycles. However, basin-scale diversity patterns and environmental drivers remain poorly understood. In this study, we applied a machine-learning framework to model the diversity of marine prokaryotic communities across the Mediterranean Sea. Diversity was quantified using the Shannon Diversity Index (SDI) derived from 16S rRNA gene sequencing. The in situ dataset included ~600 samples collected year-round from 2001 to 2023 at coastal and open-water sites, providing broad temporal coverage and multisite spatial sampling. We trained an XGBoost model using satellite-derived and modeled oceanographic variables matched to the SDI observations. The model achieved robust predictive performance (R2 = 0.78 for training and 0.70 for testing, with RMSE = 0.31 and MAPE = 0.05 across both) and captured broad basin spatial and seasonal patterns in prokaryotic community diversity, with greater uncertainty in less-represented regions. Diversity was highest in nutrient-rich coastal areas and during winter mixing, and lowest in summer-stratified or oligotrophic waters. SHAP analysis identified photoperiod as the most significant predictor, underscoring the central role of seasonal light cycles in shaping prokaryotic community diversity. Other predictors exhibited significant season- and region-dependent effects, each contributing positively within specific environmental thresholds. Climatological diversity maps revealed consistent spatiotemporal patterns, highlighting a notable west-to-east decrease in diversity and coastal hotspots. These results demonstrate that machine learning can identify major environmental drivers of prokaryotic diversity and upscale discrete observations to basin-wide predictions. This approach is transferable to other planktonic groups and supports scalable ecosystem monitoring across environmental gradients.
Keywords: prokaryotic community diversity, machine learning, Mediterranean Sea, omics, Shannon diversity index, remote sensing
Published in DiRROS: 04.05.2026; Views: 124; Downloads: 84
Full text (11,44 MB)
This document has many files! More...

Keywords: eye-tracking, machine learning, mild cognitive impairment, validation, reading characteristics
Published in DiRROS: 20.04.2026; Views: 118; Downloads: 99
Full text (1,68 MB)
This document has many files! More...

Abstract: Background: Ovarian cancer is the deadliest gynecological malignancy, largely due to the advanced stage at diagnosis in most patients. This study investigates whether systemic steroids can serve as biomarkers to distinguish malignant ovarian tumors from non-malignant adnexal masses. Methods: This prospective, single-center observational study included 99 women with adnexal masses who underwent surgery between December 2021 and February 2025. Preoperative serum levels of 17 steroid hormones, including androgens, 11-oxyandrogens, glucocorticoids, and mineralocorticoids, were quantified using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Machine learning was employed to assess the diagnostic potential of these steroids in distinguishing ovarian cancer (n = 43) from non-malignant adnexal masses (n = 56). Results: Patients with ovarian cancer had lower levels of 11β-hydroxy-testosterone (11OHT), 11-keto-testosterone (11KT), and testosterone compared to controls. Using stepwise feature selection, we developed two diagnostic models incorporating three 11-oxyandrogens (11KT, 11OHT, and 11β-hydroxy-androstenedione), patient age, and either cancer antigen 125 (CA-125) or human epididymis protein 4 (HE4) for distinguishing malignant from non-malignant adnexal masses. The model including CA-125 achieved AUC of 0.907, 88.9% sensitivity and 82.0% specificity, while the model including HE4 achieved AUC of 0.911, 94.4% sensitivity and 77.3% specificity as evaluated by cross-validation. Both models significantly outperformed CA-125, HE4, and the Risk of Ovarian Malignancy Algorithm (ROMA) index alone. Conclusion: Patients with ovarian cancer exhibit distinct steroid profiles compared to those with non-malignant adnexal masses. If validated, the models could enhance diagnosis, reducing unnecessary surgeries for benign conditions while ensuring timely treatment for ovarian cancer, particularly when conventional biomarkers are inconclusive.
Keywords: adnexal masses, diagnostic models, machine learning, ovarian cancer, steroids
Published in DiRROS: 10.04.2026; Views: 164; Downloads: 104
Full text (1,60 MB)
This document has many files! More...

Abstract: Background Immune dysregulation, including autoimmunity, autoinflammation, allergy, and malignancy predisposition, adds significant disease burden in primary immune disorders (PID) and inborn errors of immunity (IEIs). Objective We evaluated whether the 5-graded immune deficiency and dysregulation activity (IDDA2.1) score, encompassing 21 organ involvement and disease burden parameters, supports diagnosis across a wide spectrum of IEIs. Methods From April 2022 to November 2024, collaborators from 84 centers collected 1,043 IDDA score datasets from 825 patients across 89 IEIs (17 disorders with ≥10 patients each; range, 1-196 per IEI), including 177 scores from 141 treated patients. Supervised machine learning models ( k -nearest neighbors, support vector machine, logistic regression, random forest) classified patients into disease groups and ranked corresponding predictive features, while unsupervised uniform manifold approximation and projection (UMAP) visualized disease-specific clustering. Results Feature analysis reflected clinicians’ recognition of IEI patterns and confirmed internal IDDA score consistency. Phenotype profiles in treated patients remained informative, inversely reflecting anticipated treatment-dependent phenotype amelioration. UMAP effectively distinguished IEIs by IDDA2.1 profiles. Genetic disorder prediction achieved 73% overall accuracy, 70% for the correct monogenic IEI, and 93% within the top 3 predictions; classification reached 43% for IEI–International Union of Immunological Society categories and 59% for 12 “cardinal” IEIs (25 genes). Conclusions Random forest feature importance analysis can inform targeted clinical screening for key disease manifestations. The top 3 prediction approach demonstrates diagnostic potential, but improved accuracy will require larger, globally shared datasets. Small sample sizes for rare diseases highlight the necessity of broader collaboration to enhance AI-assisted clinical decision-making in the future.
Keywords: inborn error of immunity, IEI, primary immune regulatory disorder, PIRD, phenotype-driven disease classification, interoperable patient data, immune deficiency and dysregulation activity (IDDA) score, artificial intelligence, AI, unsupervised and supervised machine learning, ML, primary immune disorder, PID
Published in DiRROS: 08.04.2026; Views: 167; Downloads: 126
Full text (2,83 MB)
This document has many files! More...

Keywords: machine learning, deep learning, artificial intelligence, diabetes management, challenges
Published in DiRROS: 09.03.2026; Views: 247; Downloads: 159
Full text (636,37 KB)
This document has many files! More...