Using Machine Learning to Detect Factors That Affect Homocysteine in Healthy Elderly Taiwanese Men.

Area of Science:

• Biomedical research
• Medical informatics
• Gerontology

Background:

• Homocysteine (Hcy) is vital for physiological processes; elevated levels correlate with cardiovascular and neurological issues.
• Traditional statistical methods have limitations in identifying factors influencing Hcy.
• Machine learning (ML) offers advanced capabilities for medical research and prediction.

Purpose of the Study:

• To identify key factors influencing Hcy levels in healthy elderly Taiwanese men.
• To compare the predictive accuracy of four ML methods against multiple linear regression (MLR).
• To enhance Hcy prediction accuracy and provide insights into its determinants.

Main Methods:

• Employed four ML methods: random forest (RF), stochastic gradient boosting (SGB), eXtreme gradient boosting (XGBoost), and elastic net (EN).
• Analyzed 33 parameters in 468 healthy elderly men.
• Utilized MLR as a benchmark and assessed model performance using SMAPE, RAE, RRSE, and RMSE.

Main Results:

• All ML methods exhibited superior accuracy with lower prediction errors compared to MLR.
• C-reactive protein (CRP) was identified as the primary factor influencing Hcy levels.
• Other significant factors included GPT, WBC, LDH, eGFR, and sport volume (SV).

Conclusions:

• ML techniques significantly outperformed MLR in predicting Hcy levels in the studied population.
• CRP emerged as the most critical determinant of Hcy levels.
• GPT/ALT, WBC, LDH, and eGFR were also identified as important predictors of Hcy.