Fig. 6: Jaccard index distributions for high genetic risk determined by trait-specific Best(Pubn) vs. ChronoAdd(Pubn) from five most recent publications across genetically predicted ancestry groups.
From: Instability of high polygenic risk classification and mitigation by integrative scoring
The distribution of Jaccard indices of similarity calculated for pairs of classification of top 10% of risk determined by trait-specific polygenic scores from the last five publications across genetically inferred ancestry groups for each trait: coronary artery disease (CAD), type 2 diabetes mellitus (T2DM), and major depressive disorder (MDD). Genetically inferred ancestry based on the k-nearest neighbor approach as European (EUR), African (AFR), admixed American (AMR), and East Asian (EAS). Jaccard index distribution calculated in group EUR for traits A CAD, B T2DM, and C MDD; in group AMR for traits D CAD, E T2DM, and F MDD; in group AFR for traits G CAD, H T2DM, and I MDD; and in group EAS for traits J CAD, K T2DM, and L MDD, respectively Best(Pubn) indicates the single best trait-specific PRS from each of the last five publications, ordered by date of publication. ChronoAdd(Pubn) indicates polygenic scores generated using the PRSmix framework wherein all scores from each Pubn were combined with all the scores from all other Pubn published before it using PRSmix.
