Extended Data Fig. 5: Comparison between WGS-based and imputation-based GWAS.
From: Estimation and mapping of the missing heritability of human phenotypes
(a) Number of independent associations detected for different MAF bins and data type (WGS or imputation panels), relative to the number of associations detected using WGS data. (b) Proportion of trait-associated WGS variants with no imputed associations detected around them. Those are denoted “zero-density”. The proportion of zero-density WGS associations was calculated by varying the window size around them (x-axis) and separately for common and rare variants associations. For the largest window size (1000 kb), we indicate the actual number of zero-density WGS associations missed by imputation-based GWAS. (c) Example of zero-density common variant (MAF = 45%): intronic variant with splicing effect significantly associated with waist-to-hip ratio (WHR) in WGS GWAS but missed by imputation-based GWAS. (d) Example of zero-density rare variant (MAF = 0.8%): pathogenic SNP (PrimateAI 3D percentile score of 0.78) downstream TINF2 associated with telomere length (TELO) in WGS-based GWAS and missed by imputation-based GWAS. A high-level comparison of fine-mapping resolution between WGS and imputation is shown in Extended Data Fig. 6.
