Functional Annotations-Informed Whole Genome Sequence Analysis Identifies Novel Rare Variants for AD in the Alzheimer’s Disease Sequencing Project

Background: Incorporating functional annotations improves power to identify rare variants (RV) in the analysis of whole genome sequencing (WGS) association studies. We incorporated Alzheimer’s Disease (AD)-specific annotations based on partitioning heritability into the variant-Set Test for Association using Annotation information (STAAR-O) framework to identify RVs associated with AD in the diverse sample of the Alzheimer’s Disease Sequencing Project (ADSP). Method: We performed WGS association analyses in a sample of 4,074 individuals (1,668 AD cases; 2,406 controls) sequenced as part of the ADSP Discovery Extension Phase. We first estimated heritability of all single nucleotide variants in functional annotation categories using the GCTA tool embedded in FunSPU. For each annotation category, a global weight was derived based on the partitioned AD heritability. Among weighted functional scores, we selected the top 12 to leverage the most informative functional annotations for incorporation into the STAAR-O test. ² We performed agnostic region-based association analysis using sliding windows, defined as 2kb in length with a skip length of 1kb. Association tests were performed with RV (minor allele frequency <1%), adjusting for sex, sequencing center, platform, study, 4 principal components, and a genetic relatedness matrix. Replication of significant associations was carried out in an independent sample of 10,083 individuals (5,717 AD cases, 4,366 controls) sequenced as part of the ADSP Follow-Up Study and using the same analytical models. Result: Out of 2.65 million 2-kb overlapping windows with a total minor allele count >10, two non-consecutive windows on chromosome 17 were associated with AD (P<5×10 ⁻⁸) (Figures). Both windows associations were replicated in the independent sample (P = 0.003 and 0.016). The top variant of one significant window was rs534148850 (MAF = 0.0005, P = 5.55×10 ⁻⁹) located downstream of PLEKHM1P1, a pseudogene, and MIR4315-2, a microRNA with predicted targets enriched in apoptosis pathways. The top variant of the other window was rs532055552 (MAF = 0.005, P = 6.20×10 ⁻⁹) located downstream of CEP112, a coiled domain-containing protein involved in the regulation of gamma-aminobutyric acid A receptor surface expression. Conclusion: Incorporating AD-relevant functional annotations to a powerful RV association framework, we discovered and replicated two novel genetic regions harboring RV associated with AD.