Fig. 2: Data flow in the major analysis stages and evidence for association.
a Association testing was performed in three stages: in stage 1 (red circle), we performed single variant association tests for Cameroonian discovery (n = 827) and Tanzanian re-analysis cohort (n = 884) imputed datasets filtered to include only biallelic SNVs (SNPs and INDELs) with R2 ≥ 0.6. A generalised linear mixed model was run using SAIGE93, with multiple testing correction by the Benjamini–Hochberg false discovery rate (FDR) method; stage 2 (green circle) involved a meta-analysis of two Africa-based cohorts (n = 1711), while in stage 3 (purple circle), we included summary statistics from HbF GWAS involving sickle cell anaemia cohorts of African ancestry based in the United States of America (n = 2040) to perform an overall or a global meta-analysis (n = 3751). Meta-analysis was performed by the inverse variance method of METAL95. Summary statistics from all three stages were then used to perform functional GWAS and fine-mapping. b Overview of significant loci detected per analysis unit. Green circles indicate presence while red diamond indicates absence of significant signal in the corresponding locus. Blue coloured loci represent the major known HbF-influencing loci, while the rest in black are novel loci. c The quantile-quantile (Q-Q) plots of the expected against the observed p values, as well as the genomic control inflation factors (λ) demonstrate the robustness of our association results and show that our test statistics were not inflated. d Manhattan plots showing the significant signals in the Cameroon, Cameroon-Tanzania meta-analysis, and global meta-analysis.
