# A Scoring Model Using Multi-Metabolites Based on Untargeted Metabolomics for Assessing Dyslipidemia in Korean Individuals with Obesity

## TL;DR

This study develops a metabolite-based scoring model to assess dyslipidemia in obese Korean individuals, showing promising diagnostic accuracy when combined with traditional lipid profiles.

## Contribution

The novel contribution is the development and validation of a metabolite risk score (MRS) for dyslipidemia using untargeted metabolomics.

## Key findings

- N-acetylisoputreanine-γ-lactam and eicosapentaenoic acid were identified as key metabolites for dyslipidemia.
- The MRS model achieved 85.4% accuracy in the discovery set and 76.1% in the replication set.
- Combining MRS with lipid profiles improved diagnostic accuracy to 91.8% in the discovery set and 86.0% in the replication set.

## Abstract

Background/Objectives: Metabolite risk score (MRS), which considers the collective effects of metabolites closely reflecting a phenotype, is a new approach for disease assessment, moving away from focusing solely on individual biomarkers. This study aimed to investigate a metabolite panel for dyslipidemia and verify the diagnostic efficacy of MRS on dyslipidemia. Methods: Key metabolite identification and MRS establishment were conducted in the discovery set, and MRS validation was performed in the replication set, with 50 healthy individuals and 50 dyslipidemia patients in each set. The MRS was constructed using key metabolites, identified via UPLC-MS/MS analysis, employing a weighted approach based on linear regression analysis. Results: N-acetylisoputreanine-γ-lactam and eicosapentaenoic acid were identified as key metabolites for dyslipidemia and were utilized for establishing the MRS. In addition to the MRS model, a conventional dyslipidemia diagnostic model based on lipid profiles, as well as a combined model (MRS + lipid profiles), were also established. In the discovery set, the MRS model diagnosed dyslipidemia with 85.4% accuracy. When combined with lipid profiles, accuracy improved to 91.8%. In the replication set, the MRS demonstrated diagnostic power with 76.1% accuracy, while the combined model achieved 86.0% accuracy for dyslipidemia assessment. Conclusions: The MRS alone indicated sufficient assessment power in a real-world setting, despite a slight reduction in assessment ability when validated in the replication set. At this stage, therefore, the MRS serves as an auxiliary tool for disease diagnosis. This first attempt to apply MRS for dyslipidemia may offer a foundational concept for MRS in this disease.