Risk Prediction Using Genome-Wide Association Studies on Type 2 Diabetes

Area of Science:

• Genetics
• Statistical genetics
• Computational biology

Background:

• Genome-wide association studies (GWASs) identify genetic variants for complex diseases but often yield small effect sizes, hindering risk prediction model development.
• The "large p and small n" problem, where the number of genetic markers (p) exceeds the sample size (n), is a significant challenge in genetic risk prediction.
• Penalized regression methods offer solutions for variable selection and prediction in high-dimensional genetic data.

Purpose of the Study:

• To evaluate the performance of statistical methods for binary trait prediction, specifically for type 2 diabetes risk.
• To compare the predictive accuracy of stepwise logistic regression (SLR) combined with LASSO and Elastic-Net (EN) against other combinations.
• To identify optimal statistical approaches for building robust type 2 diabetes genetic risk prediction models.

Main Methods:

• Utilized stepwise logistic regression (SLR), least absolute shrinkage and selection operator (LASSO), and Elastic-Net (EN) for variable selection and prediction.
• Developed prediction models for type 2 diabetes using genetic data from the Korean Association Resource project (Affymetrix Genome-Wide Human SNP Array 5.0).
• Assessed model performance using the area under the receiver operating characteristic curve (AUC) for both internal and external validation datasets.

Main Results:

• In internal validation, combinations of SLR with LASSO (SLR-LASSO) and SLR with EN (SLR-EN) demonstrated superior prediction accuracy.
• During external validation, the SLR-SLR and SLR-EN combinations achieved the highest predictive performance, with an AUC of 0.726.
• The study highlights the effectiveness of specific penalized regression combinations in enhancing type 2 diabetes risk prediction.

Conclusions:

• Combinations of stepwise logistic regression with LASSO or Elastic-Net are effective for type 2 diabetes genetic risk prediction.
• These methods address the "large p and small n" challenge, offering improved accuracy over traditional GWAS approaches.
• The proposed SLR-SLR and SLR-EN models show potential as powerful tools for predicting type 2 diabetes risk.