Metabolomics Model Detects Radiation-Induced Heart Disease

Researchers have developed a diagnostic model capable of identifying radiation-induced heart disease, a serious complication arising from radiotherapy for thoracic malignancies. The study, published in Cardio-Oncology, utilized untargeted metabolomics to screen plasma samples for potential biomarkers. By analyzing metabolic profiles, the team constructed machine learning models that demonstrated high accuracy in distinguishing patients with the condition from healthy controls. This approach offers a novel strategy for the early diagnosis of a condition that is currently challenging to detect due to the lack of specific biomarkers.

The study enrolled a total of 57 patients diagnosed with radiation-induced heart disease and 22 healthy controls. The research team performed untargeted metabolomic profiling of plasma samples using gas chromatography–mass spectrometry. The samples were processed to identify differential metabolites.

To identify these differential metabolites, the researchers combined multivariate statistical analysis with univariate analysis. Following the identification of these metabolites, the team conducted metabolic pathway enrichment analysis to understand the biological processes involved. The analysis revealed that the altered metabolites were primarily involved in phenylalanine, tyrosine, and tryptophan biosynthesis, as well as arginine biosynthesis. Additionally, the metabolites were involved in the tricarboxylic acid cycle and glycine, serine, and threonine metabolism.

The results of the metabolomic profiling identified 44 differential metabolites between the radiation-induced heart disease group and the healthy control group. Of these, six metabolites were found to be upregulated, while 38 were downregulated in the patients. This imbalance indicates a distinct plasma metabolic profile disturbance in individuals suffering from the condition.

To translate these metabolic findings into a practical diagnostic tool, the researchers constructed three machine learning models: Support Vector Machine, Linear Discriminant Analysis, and Random Forest. These models were built based on the identified biomarkers to evaluate their diagnostic performance. Variable importance analysis, specifically using random forest algorithms, highlighted key metabolites such as lysine and cysteine as significant contributors to the separation of the two groups.

Among the three models tested, the Random Forest diagnostic model demonstrated the best performance in the test set. It achieved an area under the curve of 0.872, with a 95% confidence interval ranging from 0.666 to 1. The model also showed a sensitivity of 1.00 and a specificity of 0.76.

Radiation-induced heart disease is a significant concern for patients undergoing thoracic radiotherapy, and the current lack of specific biomarkers makes early diagnosis difficult. By revealing distinct plasma metabolic profile disturbances, this research provides new insights into the pathophysiology of the condition. The identification of key metabolite biomarkers through variable importance analysis offers a clearer picture of the metabolic changes associated with the disease.

While the Random Forest model showed promising results, the study has certain limitations that must be considered.

Despite these limitations, the study represents a significant step forward in the development of diagnostic tools for radiation-induced heart disease.

Subject of Research: Metabolomics-based diagnostic model for radiation-induced heart disease

Article Title: Biomarker screening and diagnostic model construction for radiation-induced heart disease based on metabolomics

Article References: Niu, R., Wang, X., Zhou, W., Zhang, L., Jia, Q., Yang, W., Bao, T., Xu, Z., & Lin, X. (2026). Biomarker screening and diagnostic model construction for radiation-induced heart disease based on metabolomics. Cardio-Oncology. https://doi.org/10.1186/s40959-026-00565-0

Image Credits: AI Generated

DOI: 10.1186/s40959-026-00565-0

Keywords: radiation-induced heart disease, metabolomics, biomarkers, machine learning, diagnostic model, plasma profiling, cardio-oncology, gas chromatography, mass spectrometry, early diagnosis, Biomarker, screening