Risk Factors for Gout in Taiwan Biobank: A Machine Learning Approach.

Area of Science:

• Computational biology
• Medical informatics
• Genetics

Background:

• Gout is a common inflammatory arthritis caused by hyperuricemia.
• Accurate prediction of gout risk is crucial for early intervention and management.

Purpose of the Study:

• To assess gout risk in the Taiwan Biobank population using machine learning algorithms.
• To identify critical risk factors for gout.
• To evaluate the predictive performance of various machine learning models for gout.

Main Methods:

• Analysis of data from 88,210 individuals in the Taiwan Biobank.
• Application of five machine learning models: Bayesian Network, Random Forest, Gradient Boosting, Logistic Regression, and Neural Network.
• Propensity score matching for gender and age to create a balanced sample of 38,676 individuals (19,338 gout cases, 19,338 controls).
• Evaluation using an 80% training and 20% test set split.

Main Results:

• Uric acid and gender were identified as the most significant risk factors for gout.
• Random Forest (RF) and Gradient Boosting (GB) models exhibited high performance.
• RF achieved an Area Under the Curve (AUC) of 0.986-0.987, sensitivity of 0.945-0.947, and specificity of 0.998-0.999.
• GB showed comparable results with AUC around 0.987-0.988, sensitivity of 0.944-0.950, and specificity of 0.995-0.999.
• Both RF and GB achieved high F1 scores (0.971-0.972).

Conclusions:

• Random Forest and Gradient Boosting models demonstrate exceptional accuracy in predicting gout.
• The integration of genetic data with clinical factors can further enhance gout prediction.
• These findings support the clinical application of machine learning for improved gout risk assessment.