Methods for estimating insulin resistance from untargeted metabolomics data.

Area of Science:

• Metabolomics
• Biostatistics
• Human Health

Background:

• Insulin resistance (IR) is linked to numerous complex diseases.
• Current methods for measuring IR, such as insulin sensitivity index (SI) and homeostatic model assessment of insulin resistance (HOMA-IR), are invasive, costly, and time-consuming.
• There is a need for more accessible and efficient methods to assess IR.

Purpose of the Study:

• To develop and validate estimation models for key measures of insulin resistance (SI and HOMA-IR) using metabolomics data.
• To assess the feasibility of using metabolomics data combined with clinical factors for IR estimation.
• To evaluate the performance of these models in diverse populations, including Mexican Americans and African Americans.

Main Methods:

• Utilized Least Absolute Shrinkage and Selection Operator (LASSO) and Elastic Net regression.
• Integrated 1274 metabolites with clinical data (age, sex, BMI) to build estimation models.
• Employed various data transformation techniques (inverse normal, standardization, Box Cox) for metabolite data.
• Validated models across different recruitment sites and ethnic groups (Mexican Americans and African Americans).

Main Results:

• Developed robust estimation models for SI and HOMA-IR using metabolomics and clinical data.
• Achieved high correlations between estimated and empirical SI in testing datasets (r²=0.77 in Mexican Americans, r²=0.74 in African Americans).
• Demonstrated consistent associations between estimated SI and key clinical markers like BMI, LDL cholesterol, and triglycerides.

Conclusions:

• Successfully developed a novel method for estimating insulin resistance using metabolomics data.
• The developed models show strong predictive power and clinical relevance.
• This approach has significant potential for broad application in biomedical research and clinical practice for assessing insulin resistance.