Machine learning-based differentiation of uni- and bilateral primary aldosteronism using circulating microRNA combinations

Balint Vekony; Gabor Nyirő; Zoltan Herold; Janos Fekete; Filiipo Ceccato; Sven Gruber; Lydia Kurzinger; Mirko Parasiliti-Caprino; Nikolette Szucs; Balint Szeredas; Siti Khadija Syed Mohammed Nazri; Vanessa Fell; Mohamed Bassiony; Elena A B Aisha Azizan; Irina Bancos; Felix Beuschlein; Peter Igaz

doi:10.1530/endoabs.99.OC11.4

OC11.4

Prev Next

Section Contents Cite

Endocrine Abstracts (2024) 99 OC11.4 | DOI: 10.1530/endoabs.99.OC11.4

ECE2024 Oral Communications Oral Communications 11: Adrenal and Cardiovascular Endocrinology | Part II (6 abstracts)

Machine learning-based differentiation of uni- and bilateral primary aldosteronism using circulating microRNA combinations

Bálint Vékony ^1,2 , Gábor Nyirő ^1,2,3 , Zoltán Herold , János Fekete ⁴ , Filiipo Ceccato ⁵ , Sven Gruber ⁶ , Lydia Kürzinger ⁷ , Mirko Parasiliti-Caprino ⁸ , Nikolette Szücs ^1,2 , Bálint Szeredás ^1,2 , Siti Khadija Syed Mohammed Nazri ⁹ , Vanessa Fell ¹⁰ , Mohamed Bassiony ¹⁰ , Elena A B Aisha Azizan ⁹ , Irina Bancos ¹⁰ , Felix Beuschlein ⁶ & Péter Igaz ^1,2

Views

Author affiliations

¹Semmelweis University, Department of Internal Medicine and Oncology, Budapest, Hungary; ²Semmelweis University, Department of Endocrinology, Budapest, Hungary; ³Semmelweis University, Department of Laboratory Medicine, Budapest, Hungary; ⁴Semmelweis University, Department of Bioinformatics, Budapest, Hungary; ⁵University of Padova, Endocrinology Unit, Department of Medicine, Padova, Italy; ⁶University Hospital and University of Zurich, Department of Endocrinology, Diabetology and Clinical Nutrition, Zurich, Switzerland; ⁷University of Würzburg, Department of Internal Medicine I, Division of Endocrinology and Diabetes, Würzburg, Germany; ⁸University of Turin, Endocrinology, Diabetes and Metabolism, Department of Medical Sciences, Turin, Italy; ⁹The National University of Malaysia, Endocrine Unit, Department of Medicine, Kuala Lumpur, Malaysia; ¹⁰Mayo Clinic, Division of Endocrinology, Diabetes, Metabolism and Nutrition, Department of Internal Medicine, Rochester, United States

Background: Primary aldosteronism (PA) is the most prevalent cause of secondary hypertension. Its two main clinical forms are unilateral adenoma (APA) and bilateral hyperplasia (BAH), which require markedly different medical treatments, so differentiating between the two is of utmost clinical importance. The current gold standard method for this is adrenal vein sampling (AVS), the application of which is hindered by limited availability and high skill requirements.AimOur goal was to identify circulating microRNAs – or their combinations – which enable differentiation between the two most prevalent aetiologies of PA from a peripheral blood sample.

Methods: MicroRNA specific sequencing was performed on an Illumina platform, using EDTA coagulated blood samples taken during AVS, from 18 patients (10 uni-, and 8 bilateral). First, plasma samples from both adrenal veins were evaluated. Bioinformatical analysis applying the DeSeq2 algorithm was used to evaluate the differences in expression; and a neural network model, tasked to identify the most fit individual and groups of microRNAs for differentiation was used. The microRNAs comprising the five best performing models were then validated using reverse transcription real-time PCR, on 90 samples, including right and left AVS and peripheral plasma samples from 30 patients (15 uni-, and 15 bilateral). The qPCR results were then re-analysed using the same neural network. Finally, the model was validated on an independent peripheral plasma sample group from 84 patients (42 uni-, and 42 bilateral).

Results: Based on the qPCR results, miRNA abundance shows a non-significant decrease on the periphery compared to the adrenal vein samples. 10 neural network models comprised of 4 to 10 microRNAs were able to differentiate BAH and APA using peripheral plasma samples with an accuracy above 85%, with the best model consisting of 6 miRNAs having a specificity of 87.91%, a sensitivity of 86.3%, and an AUC value of 87.1%. The inclusion of clinical parameters into the models, such as imaging results, aldosterone, renin and potassium plasma levels and BMI did not meaningfully alter the performance of the model.

Conclusion: Circulating microRNA-based differentiation between uni-, and bilateral PA could prove to be a reliable diagnostic method. Should the diagnosis of BAH be established, medical treatment may begin immediately, with no further tests required to establish localization. This could prove highly beneficial for decreasing the health and financial costs of patients and providers alike.