OC 18.2 - A signature of platelet reactivity in complete blood count scattergrams identifies genes with thrombotic disease associations

Background: Genetic studies of platelet reactivity (PR) offer a hypothesis-free approach for antiplatelet drug target discovery. Unfortunately, PR is technically challenging to measure accurately. PR is typically measured with a platelet function assay or by light transmission aggregometry. These assays are time consuming and difficult to standardize. Consequently, genetic studies of PR have been hampered by small (N < 4,000) sample sizes, identifying, in aggregate, only 15 distinct PR-associated loci. A higher throughput, standardised signature of PR using platelet scattergrams
from Sysmex haematology analysers would enable biobank-scale (N >10,000) genetic studies of PR.

Aims: Our aims were to (1) train a predictor of flow cytometry-measured PR from Sysmex platelet channel scattergrams in the Cambridge Platelet Function Cohort (PFC); (2) perform a genome-wide association study (GWAS) of this predicted phenotype in 29,806 blood donors in INTERVAL; (3) build a genetic score of PR and (4) assess whether this genetic score of predicted PR is associated with health outcomes in UK Biobank.

Methods: We extracted features capturing the variation between Sysmex scattergrams. We fit a Lasso regression using data from the 533 participants in the PFC for which both Sysmex scattergrams and flow-derived platelet activation measurements were available. We predicted PR in response to ADP in INTERVAL from Sysmex data, performed a GWAS of predicted PR and built a genetic score. We performed survival analyses in UK Biobank to test for associations between the score and 524 health outcomes.

Results: We identified 21 distinctly associated loci, only 6 of which were previously known. The score of PR was associated with thrombosis-related health outcomes.

Conclusion(s): Sysmex scattergrams contain a signature of PR, allowing biobank scale studies of PR. We linked a genetic score of PR to thrombosis in a population cohort. Our approach provides a blueprint for studying the determinants of hard-to-measure but biologically important traits by phenotype imputation.