Definition of high-risk type 1 diabetes HLA-DR and HLA-DQ types using only three single nucleotide polymorphisms

Area of Science:

• Immunogenetics
• Computational Biology
• Diabetes Research

Background:

• Type 1 diabetes (T1D) risk assessment relies on human leukocyte antigen (HLA) typing, specifically HLA DRB1 and DQB1 alleles.
• Certain HLA types, such as DR3 and DR4 in combination with DQ8, confer the highest risk for T1D development.
• Existing HLA typing methods are costly and time-intensive, hindering widespread risk assessment.

Purpose of the Study:

• To identify the minimum number of single nucleotide polymorphisms (SNPs) required for rapid and accurate determination of T1D-relevant HLA-DR and HLA-DQ types.
• To develop a cost-effective method for predicting high-risk T1D genotypes, including distinguishing DR4-DQ8 and DR4-DQB1*03:01.

Main Methods:

• Analysis of 19,035 SNPs across 10,579 subjects from the Type 1 Diabetes Genetics Consortium (discovery and validation sets).
• Development of a novel machine learning algorithm to select a minimal set of SNPs for HLA type prediction.
• Validation of SNP reliability using 10-fold cross-validation.

Main Results:

• A machine learning approach identified as few as three SNPs capable of accurately defining HLA-DR and HLA-DQ types relevant to T1D.
• The developed method achieved high accuracy (99.3%), with an area under the curve of 0.997, high true-positive rates (>0.99), and very low false-positive rates (<0.001).
• The selected SNPs reliably predicted T1D-associated HLA types, including high-risk genotypes.

Conclusions:

• A rapid, cost-effective method using a minimal set of SNPs can accurately predict T1D-associated HLA-DR/DQ types.
• This SNP-based approach offers a significant improvement over current, more expensive, and time-consuming HLA typing methods for T1D risk assessment.
• The findings pave the way for more accessible and efficient T1D genetic risk screening.