Machine Learning-Based Method for Obesity Risk Evaluation Using Single-Nucleotide Polymorphisms Derived from...

Area of Science:

• Genetics and Genomics
• Computational Biology
• Metabolic Disease Research

Background:

• Obesity is a significant risk factor for numerous metabolic diseases.
• Next-generation sequencing (NGS) provides comprehensive genome-wide genetic data, including single-nucleotide polymorphisms (SNPs).
• Interpreting complex SNP data for clinical applications presents challenges.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting obesity risk based on SNP data.
• To identify informative SNPs and assess their effectiveness in obesity prediction.
• To compare the performance of different ML algorithms for obesity risk assessment.

Main Methods:

• Utilized clinicopathological data, including 130 SNPs, sex, and age, from 139 individuals.
• Applied feature selection methods like stepwise multivariate linear regression (MLR) and decision trees (DT) to identify key SNPs.
• Developed obesity prediction models using Support Vector Machine (SVM), k-nearest neighbor, and DT algorithms.
• Evaluated model performance using fivefold cross-validation, assessing accuracy, sensitivity, and specificity.

Main Results:

• Nine informative SNPs (rs10501087, rs17700144, rs2287019, rs534870, rs660339, rs7081678, rs718314, rs9816226, rs984222) were selected using stepwise MLR.
• The Support Vector Machine (SVM) model demonstrated superior performance compared to other classifiers.
• The SVM model achieved 70.77% accuracy, 80.09% sensitivity, and 63.02% specificity.

Conclusions:

• Selected SNPs are effective in detecting obesity risk.
• Machine learning-based approaches offer a feasible method for preliminary analysis of obesity's genetic characteristics.
• This study highlights the potential of integrating genetic data and ML for personalized obesity risk assessment.