Metabolomics biomarkers for predicting renal function decline

2020 Fiscal Year Final Research Report

• Project/Area Number: 18K17394
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 58030:Hygiene and public health-related: excluding laboratory approach
• Research Institution: Keio University
• Principal Investigator: HARADA Sei 慶應義塾大学, 医学部(信濃町), 講師 (10738090)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Keywords: 腎機能低下 / メタボローム / 機械学習 / 予防医学 / 慢性腎臓病

Outline of Final Research Achievements

Renal function measurements (serum creatinine, serum cystatin C, and urine albumin) and plasma and urine metabolomics were performed on 1,672 participants aged 60-74 years at the beginning of the study. The same participants were also evaluated in the same way after 6 years.
A machine learning method (OPLS-DA) was used to predict the decline in renal function over 6 years, and a more accurate prediction model was constructed by using plasma and urine metabolome in addition to classical renal function indicators. Furthermore, another machine-learning approach (SVM) was used to create ROC curves, which were most accurate when the top five variables including three metabolites were selected, with a good performance of AUC 0.904 (95%CI 0.871-0.944).

Free Research Field

メタボロミクス疫学

Academic Significance and Societal Importance of the Research Achievements

慢性腎臓病は、高齢および糖尿病が大きなリスク因子であることから、高齢社会の進展、また糖尿病の有病者の増加する現代の我が国においては、公衆衛生学上の重要な課題となっているが、近い将来の腎機能低下を予測することは困難なことも多い。
本研究では、既知のバイオマーカーに加えて、血漿メタボロームおよび尿中メタボロームを測定することで、将来の腎機能低下をより早く正確に捉えることができる可能性が示唆され、予防医療への応用可能性が示された。