Tlalpan 2020 Case Study: Enhancing Uric Acid Level Prediction with Machine Learning Regression and Cross-Feature...

Area of Science:

• Metabolic health and disease prediction
• Biomarker analysis and machine learning applications

Background:

• Uric acid is a metabolic byproduct with dual roles: antioxidant at physiological levels and a contributor to diseases like gout and cardiovascular issues when elevated.
• Hyperuricemia is linked to metabolic disorders, including hypertension and insulin resistance, underscoring the need to understand its regulation.

Purpose of the Study:

• To predict uric acid levels using machine learning algorithms.
• To identify key clinical, anthropometric, lifestyle, and nutritional variables associated with hyperuricemia.

Main Methods:

• Application of Boosted Decision Trees (Boosted DTR), eXtreme Gradient Boosting (XGBoost), and Categorical Boosting (CatBoost) models.
• Utilized Shapley Additive Explanations (SHAP) for variable importance.
• Employed feature engineering and cross-feature selection to enhance model performance, evaluated by MSE, RMSE, and R².

Main Results:

• XGBoost excelled with anthropometric/clinical data; CatBoost identified nutritional risk factors.
• Distinct gender-specific predictive profiles for uric acid levels were observed.
• Men's levels influenced by renal function, lipids, and paternal history; women's by metabolic/cardiovascular markers and lifestyle.

Conclusions:

• Machine learning models effectively predict uric acid levels and identify key determinants.
• Findings reveal distinct metabolic, nutritional, and lifestyle factors influencing uric acid in men and women.
• Supports targeted public health strategies for hyperuricemia prevention based on gender-specific insights.