Genetics
Integration of functional genomics in variant-to-gene annotation for Alzheimer’s Disease pathway-specific polygenic risk score construction Katrina Bazemore* Katrina Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore Bazemore University of Pennsylvania
Pathway-specific polygenic risk scores (pathway PRS) measure aggregated genetic liability to disease within biological pathways, better representing the heterogeneity of individual-level genetic risk composition than global PRS. Pathway PRS measure liability across single nucleotide variants annotated to genes in the pathway of interest. In application, most variant-to-gene annotation is based on variant position with respect to gene boundaries. This is error-prone for intergenic and intronic variants, which can regulate gene expression at long distances and are a large proportion of risk variants for Alzheimer’s Disease (AD). We hypothesized that integrating chromatin interaction and expression quantitative trait locus (eQTL; variants associated with gene expression) data derived from adult brain tissue (nbrains=3 and 1,387) will improve AD pathway PRS performance. We trained 20 AD pathway PRS with clumping and thresholding on positional and integrative variant-to-gene annotation using ten repeated 80:20 holdouts in the UK Biobank. AD cases had validated ICD9/10 codes indicating AD (true cases) or reported parental AD/dementia (proxy cases) (ncases=41,632; ncontrols=286,904); a sensitivity analysis used only true cases (ncases=2,917; ncontrols=256,179). The integrative strategy annotated fewer variants to pathways than positional, though the linkage disequilibrium-independent variant count was similar across strategies (median differences: -46% and 5%). Most pathway PRS had greater incremental R2 under integrative annotation for both outcome definitions (Figure 1). The top pathway PRS included genes from 4 Gene Ontology (GO) amyloid-related biological processes (GO:0034249, GO:1902991, GO:1902430, GO:1902992). Integrative variant-to-gene annotation improved AD pathway PRS performance with fewer variants overall. This strategy uses publicly available data and can be implemented to improve power and precision in pathway PRS applications across outcomes.
