Translating autoimmune genetic risk scores into improved prediction and classification of disease

The clinical translation of established genetic risk for type 1 diabetes is hindered by a lack of standardisation and understanding of disease heterogeneity. This thesis aims to address these issues by utilising genetic risk scores (GRS) to enhance type 1 diabetes prediction, differential diagnosis, and mechanistic classification. The lack of genetic risk score standardisation typically arises due to their complex generation and use of proxies within missing data. In chapter 2, we address measurement inconsistency by developing the GRS2x model and the open-source PRSedm Python package, enabling reproducible and standardised polygenic risk calculation across array and whole-genome sequencing data. GRS2x demonstrated strong discriminatory performance in European populations, with utility in trusted research environments and ability to generate scores for other traits. Current autoantibody screening recommendations for type 1 diabetes would not be applicable for individuals diagnosed at <2 years. Therefore, we investigated the clinical utility of genetic risk scores in identifying these individuals who would benefit from follow up monitoring for type 1 diabetes and also for recommending individuals for monogenic testing. In chapter 3, the GRS was shown to differentiate between monogenic autoimmune diabetes and polygenic type 1 diabetes. This finding provides a rationale for using a GRS threshold as a cost-effective diagnostic filter to guide targeted monogenic sequencing. In chapter 4, we found a strong inverse correlation with age of onset and genetic risk, highlighting how genetic risk scores can be applied in screening scenarios to not only identify individuals for autoantibody surveillance but also identify those at extremely high risk of developing rapid autoimmunity. The suggestion of endotypes within type 1 diabetes has been proposed based on previous findings of differences in first-developing autoantibodies, age at onset and immune cell islet infiltration. In chapter 5, we evaluated this further by performing large-scale GWAS stratified by HLA haplotype as a proxy for autoantibody status. We revealed a moderate genetic correlation between these groups, consistent with the suggestion of distinct endotypes. Mechanistic analysis identified unique immunological signatures: DR4-T1D risk was more enriched in T-cell regulatory elements, while DR3-T1D risk was uniquely enriched in mast cell cis-regulatory elements. These findings provide novel genetic evidence for divergent autoimmune pathways and offer specific, testable targets for stratified therapeutic development. This work advances the classification of type 1 diabetes by providing standardised tools and evidence-based thresholds for diagnosis and prediction. By integrating GRS with HLA-driven mechanistic insights, this thesis contributes to research which aims for equitable, personalised prediction and targeted intervention of type 1 diabetes.<p></p>

Author dc:creator

• Amber Luckett (21061592)

• Genetics
• Type 1 diabetes
• GRS
• Monogenic autoimmune diabetes
• HLA
• Genomics

Statement dc:rights

• All rights reserved

Identifier

10779/exe.31685974.v1

OAI identifier oai:identifier

oai:figshare.com:article/31685974