Kuo, Chieh-WenChieh-WenKuoHUI-AN CHENRAI-HSENG HSUWu, Chao-SzuChao-SzuWuHsu, ChingChingHsuLee, Ming-JenMing-JenLeeFENG-JUNG YANGLin, Ru-JenRu-JenLinYIN-HSIU CHIENHSUEH-WEN HSUEHPI-CHUAN FANWEN-CHIN WENGTA-CHING CHENCHIH-CHAO YANGWANG-TSO LEEWUH-LIANG ​​HWUNI-CHUNG LEE2025-08-042025-08-042025-09https://scholars.lib.ntu.edu.tw/handle/123456789/730942Diagnosing mitochondrial diseases remains challenging because of the heterogeneous symptoms. This study aims to use machine learning to predict mitochondrial diseases from phenotypes to reduce genetic testing costs. This study included patients who underwent whole exome or mitochondrial genome sequencing for suspected mitochondrial diseases. Clinical phenotypes were coded, and machine learning models (support vector machine, random forest, multilayer perceptron, and XGBoost) were developed to classify patients. Of 103 patients, 43 (41.7%) had mitochondrial diseases. Myopathy and respiratory failure differed significantly between the two groups. XGBoost achieved the highest accuracy (67.5%). In conclusion, machine learning improves patient prioritization and diagnostic yield.enMachine learningMitochondrial diseasesPhenotype[SDGs]SDG3journal article10.1016/j.mito.2025.10206140541686