Improving T2D machine learning-based prediction accuracy with SNPs and younger age.

Area of Science:

• Genomics
• Machine Learning
• Diabetes Research

Background:

• Type 2 Diabetes (T2D) poses a significant public health challenge.
• Accurate risk prediction is crucial for timely intervention.
• Machine learning (ML) models offer potential for improved T2D risk assessment.

Purpose of the Study:

• To evaluate the impact of integrating clinical and genomic data on ML model performance for T2D risk prediction.
• To compare model performance using different data combinations and age groups.

Main Methods:

• Six ML algorithms were trained and tested on a discovery dataset.
• Models were validated using the UK Biobank dataset.
• Performance was assessed with clinical data alone, combined data, and in age-specific cohorts.

Main Results:

• Genomic data integration provided a modest performance boost across ML models.
• Clinical factors like family history and hypertension were key predictors.
• Specific SNPs and polygenic risk scores (PRS) improved prediction, particularly in individuals ≤55 years.
• Models achieved AUCs >91% in the UK Biobank with combined data.

Conclusions:

• Clinical factors are strong T2D predictors, but genomic data offers incremental improvement.
• Integrating genomic data enhances T2D risk prediction, especially for early detection in younger adults.
• Multi-dimensional models incorporating genomics show promise for refining T2D risk assessment.