PhD Scientific Days 2026

Budapest, 16-18 June 2026

Poster Session 1.S - Conservative Medicine

Validation of a circulating microRNA combination for lateralizing Primary Aldosteronism: A multi-center machine learning study

Name of the presenter

Vékony, Bálint

Institute/workplace of the presenter

Department of Internal Medicine and Oncology, Department of Endocrinology, Semmelweis University

Authors

Bálint Vékony1,2,3, Gábor Nyirő1,2,4, Zoltan Herold1,5, Bálint Kende Szeredás1,2, Markus Kroiss6, Chiara Grasselli7, Sven Gruber8, Tomaz Kocjan9, Piotr Kmieć10,11, Renata Świątkowska-Stodulska10,11, Michal Hoffmann12,13, Piotr Glinicki14, Alicja Duquenne14, Nikolette Szücs1,2, Judit Tőke1,2, Vanessa Fell15, Bahaa Salama15, Ismail Bin Osman16, Troy Puar Hai Kiat16,17, Irina Bancos15, Felix Beuschlein6,8,18, Martin Reincke6, Péter Igaz1,2
1: Department of Internal Medicine and Oncology, Faculty of Medicine, Semmelweis University, Budapest, Hungary
2: Department of Endocrinology, Faculty of Medicine, Semmelweis University, Budapest, Hungary
3: Department of Molecular Genetics and the National Tumor Biology Laboratory, National Institute of Oncology, Budapest, Hungary
4: Department of Laboratory Medicine, Semmelweis University, Budapest, Hungary
5: National Institute of Oncology, Budapest, Hungary
6: Medizinische Klinik und Poliklinik IV, LMU Hospital, Ludwig Maximilian University, Munich, Munich 80336, Germany
7: Hypertension Unit of Second Internal Cardiovascular Medicine, Azienda Unità Sanitaria Locale-IRCCS di Reggio Emilia, 42123 Reggio Emilia, Italy
8: Department of Endocrinology, Diabetology and Clinical Nutrition, University Hospital Zurich (USZ) and University of Zurich (UZH), Zurich, Switzerland
9: Department of Endocrinology, Diabetes and Metabolic Diseases, University Medical Center Ljubljana, Ljubljana, Slovenia
10: Department of Endocrinology and Internal Diseases, Medical University of Gdańsk, Gdansk, Poland
11: Department of Endocrinology and Internal Medicine, University Clinical Center, Gdansk, Poland
12: Department of Hypertension and Diabetology, Medical University of Gdańsk, Poland
13: Department of Hypertension and Diabetology, University Clinical Center, Gdansk, Poland
14: EndoLab Laboratory, Department of Endocrinology, Centre of Postgraduate Medical Education, Warsaw, Poland
15: Division of Endocrinology, Diabetes, Metabolism and Nutrition, Department of Internal Medicine, Mayo Clinic, 200 First Street SW, Rochester, MN 55905 USA
16: Changi General Hospital, Singapore, Singapore
17: Duke-NUS Medical School, Singapore, Singapore
18: The LOOP Zurich - Medical Research Center, Zurich, Switzerland

Text of the abstract

Background: We previously identified 6 circulating microRNAs (miRNAs) (miR-146a-5p, miR-24-3p, miR-130b-3p, miR-99b-5p, miR-151a-3p, miR-199a-3p) that in combination could distinguish unilateral primary aldosteronism (UPA) from bilateral adrenal hyperplasia (BAH) using neural network and deep learning approaches. Validation remains the critical bottleneck for clinical translation of PA prediction models.
Aim: To validate the clinical utility of the miRNA combination in a larger cohort of PA patients.
Methods: 249 peripheral plasma samples from patients with confirmed PA under an international collaboration of 11 expert centers were included. All patients underwent adrenal venous sampling. Plasma miRNA expression was quantified by RT-qPCR. We expanded our analytical framework to six machine learning algorithms: the original neural network and deep learning models, plus Random Forest, XGBoost, Support Vector Machine, Logistic Regression and Naive Bayes.
Results: Initial cross-center and cross-time heterogeneity prompted multiple normalization solutions, which successfully enabled model generalization across centers and cohorts. Critically, validation performance matched the training performance. On a 357 strong harmonized dataset combining our original 108 (published in 2024) and this additional 249 patient validation cohorts, the combined performance of the models yielded an area under curve (AUC) of 93.1% and regarding BAH a negative predictive value of 84.8% and an overall F1 score of 78.1%.
Conclusion: This study presents a harmonization framework for cross-center application of circulating microRNA-based lateralization of PA. The combined ensemble performance across 313 samples of 81.5% AUC represents a robust, clinically generalizable performance from this method, however further cohort expansion may increase clinical utility, thus expediting care for PA patients worldwide by the simplification of patient pathways, as PA cases confirmed as bilateral can avoid further cumbersome and expensive lateralization procedures.