P3‐248: stratification of individuals for pet amyloid positivity and alzheimer's disease risk using polygenic risk score analysis: new opportunities for clinical trial design

PET amyloid imaging and/or testing of CSF are established techniques for selection of subjects for clinical trials. The use of such tests is predicated on the observation that amyloid build up may occur many years before clinical symptoms, but represent a significant risk for future cognitive decline and development of Alzheimer's Disease. Amyloid assessment by PET may be used to stratify subjects, irrespective of the mechanism of action of the treatment. The use of polygenic risk score algorithms to stratify subjects presents the possibility of reducing failure rates using PET imaging, with associated efficiencies. Further stratification based on pathway analysis opens the possibility of better alignment between drug mechanism and trial population. A comprehensive panel of SNP DNA variants has been prepared, comprising variants associated with pathways relating to AD, whole exome analysis, and variants downstream of mTOR. All samples were genotyped on the variaTECTTM plates. For the statistics-based PRS, a training set was compiled using well clinically and biomarker-phenotyped cases and control samples and data used to identify various PRS models describing risk of Aβ-42 positivity, as assessed by PET imaging and/or CSF testing. The models were identified using a combination of methods including machine learning, a hypothesis driven naïve Baeyes approach and a hypothesis free variant selection with elastic net regularization. The mTOR-pathway modelling approach was based on theoretical knowledge of the pathways involved. The models were then tested against in various independent datasets. The genotyping data derived from ∼2,000 well-phenotyped clinical samples has been used to develop and test novel polygenic risk score (PRS) algorithms. Our results indicate that such algorithms have the potential to be deployed in order to identify and enrich amyloid-positive individuals from early symptomatic and pre-symptomatic (prodromal) cohorts. Furthermore, PRS algorithms have been identified which can be used to stratify AD risk in Apo4 non-carriers. The application of PRS algorithms can significantly reduce screening failure rates associated with the use of PET amyloid imaging in early symptomatic and pre-symptomatic subjects and offers the potential for stream-lining clinical trial recruitment. Stratification for AD risk in ApoE4 non-carriers, represents a significant step forward.