Our Height, Weight, and Family History... A Diabetes Risk Prediction Model Tailored for Koreans Has Been Developed

Overview of Developing a Machine Learning Model for Predicting Type 2 Diabetes Onset Specialized for Koreans.

[Asia Economy Yeongnam Reporting Headquarters Reporter Hwang Dooyul] A machine learning model specialized for predicting the onset of type 2 diabetes in Koreans has been developed.

Professor Lee Jeonghye’s team from the Department of Industrial Engineering at UNIST and Professor Kang Jihoon’s team from the Department of Family Medicine at Kosin University Gospel Hospital developed a machine learning model with improved predictive performance for type 2 diabetes onset based on a large-scale Korean cohort.

They developed a Genome-wide Polygenic Risk Score (gPRS) specialized for Koreans and utilized demographic information, clinical data, and metabolomic information together.

Diabetes is a common disease affecting one in six Koreans over the age of 30. It is a dangerous chronic disease that causes complications such as stroke and cardiovascular diseases, making prevention important.

The onset of diabetes is influenced by lifestyle habits including diet and genetic conditions, and research on predictive models based on such information is ongoing.

Previous studies on diabetes onset risk prediction models mainly targeted Western populations. Even when targeting East Asians, demographic information such as height, weight, and family history, or clinical data such as glycated hemoglobin (HbA1c) levels and cholesterol levels were primarily used.

As a result, there were limitations in predicting diabetes that reflected genetic and environmental factors specialized for Koreans. Therefore, the research team challenged themselves to develop a prediction model using information specialized for Koreans.

The researchers developed the prediction model based on a large-scale cohort from the Korean Genome Epidemiology Study (KoGES) collected by the National Institute of Health under the Korea Disease Control and Prevention Agency.

The cohort data, collected since 2001 for research on chronic diseases common in Koreans such as diabetes, hypertension, obesity, and metabolic syndrome, was supplemented with demographic, clinical, genetic, and environmental information to enhance the predictive performance for diabetes onset.

Seokju Han, the first author and a doctoral researcher in the Department of Industrial Engineering at UNIST, explained, “Genetic information related to type 2 diabetes onset was newly calculated as a ‘polygenic risk score’ tailored to Korean genetic characteristics and utilized in the prediction model. Environmental information was reflected through ‘metabolomics’ to complement information not explained by genetic data.”

The final developed type 2 diabetes onset prediction model showed about an 11 percentage point higher predictive performance compared to using only demographic information. It also showed more than a 4 percentage point improvement compared to using both demographic and clinical information.

Su-hyun Kim, co-first author and doctoral researcher in the Department of Industrial Engineering at UNIST, said, “The more demographic and clinical information obtained from the Korean cohort was combined with the newly developed polygenic risk score and metabolomic data, the higher the model’s predictive accuracy became.”

The newly developed model can identify the risk of diabetes onset specialized for the Korean population and provide information on contributing factors.

The research team anticipated, “If this model is utilized in clinical settings, type 2 diabetes can be effectively prevented and managed.”

Professor Lee Jeonghye stated, “It is significant that the research, which was previously centered on Western cohorts, was approached using a Korean cohort,” and added, “It is expected to be used in various follow-up studies utilizing cohort data from Asian populations.”

This study was published in eBioMedicine, a sister journal of The Lancet in the medical field, and was conducted with support from UNIST’s ‘U-K Brand Development Project (Free Innovation Research)’ and the National Research Foundation of Korea’s ‘Regional Hub Innovative Physician-Scientist Joint Research.’

Researchers at UNIST who conducted this study (from left): Researcher Kim Suhyun, Researcher Han Seokju, Professor Lee Jeonghye.

This content was produced with the assistance of AI translation services.