Supervariant identification and validation procedure. (A) Workflow of supervariant construction following a local ranking and aggregation procedure. (B) Workflow of discovery, internal validation, and external validation of supervariants. UKB White British data set is randomly split into two parts. On the first part, supervariants are constructed following the four steps in A. On the second part, the association between constructed supervariants and phenotype is validated. This discovery and internal validation procedure is repeated 10 times. Supervariants that can be discovered and validated multiple times on the UKB White British data set are regarded as reproducible supervariants. They are further validated on three external data sets with European ancestry. Then, meta-analysis is performed to combine the results. (UKB) UK Biobank, (UKBW) UKB White but non-British, (ABCD) Adolescent Brain Cognitive Development study, and (HCP) Human Connectome Project.
