Risk prediction model for chronic kidney disease using regularized regression methods

2019 Fiscal Year Final Research Report

• Project/Area Number: 19K21461
• Project/Area Number (Other): 18H06380 (2018)
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Multi-year Fund (2019); Single-year Grants (2018)
• Review Section: 0908:Society medicine, nursing, and related fields
• Research Institution: Fujita Health University
• Principal Investigator: FUJII Ryosuke 藤田医科大学, 医療科学部, 助教 (60823846)
• Project Period (FY): 2018-08-24 – 2020-03-31
• Keywords: ゲノム疫学 / 機械学習 / 正則化回帰モデル / 慢性腎臓病 / 遺伝的多型

Outline of Final Research Achievements

There has been attracted much interest in the prediction using combined information of individual's genetic variants with laboratory testing value and lifestyle habit. In this study, we developed a predictive model for chronic kidney disease (CKD) using regularized regression models based on genetic information and basic characteristics (age, gender) in a Japanese population (about 14,000 healthy people). Comparing the performance of regularized regression models with that of a conventional analysis method (linear regression model), the Lasso regression and the elastic net may construct a high-performance model with fewer variables and fewer errors. However, the improvement in performance is slight and further studies need to be included more genetic variants.

Free Research Field

疫学

Academic Significance and Societal Importance of the Research Achievements

本研究では、慢性腎臓病（CKD）をアウトカムとして予測モデルの構築に取り組んだが、他の人種もしくは疾患にも応用可能であり、個人のゲノム情報を使用したリスク予測モデル構築において、大きな可能性を秘めている研究と考えている。また、CKDを経て末期腎不全を発症すると人工透析を要することから、縦断的な検討をさらに実施することで、個人のQOLや社会生産性の向上に貢献しうると考えている。