Fig. 3: Dissection of associations between FHR-4 levels and AMD susceptibility.

a Haplotype analyses based on the AMD-associated risk allele C at rs1061170, protective allele A at rs800292, and protective CFHR3/1 deletion tagging allele T at rs12144939 and independent quantitative trait loci (QTL) for CFHR4 rs61818956, rs10494745, and rs7531555. Odds ratios (OR), confidence intervals (CI), and p-values (two-sided) were obtained using logistic regressions including AMD case/control status as the dependent variable and age, sex, and the first two genetic principal components for the IAMDGC cohort as covariates. Bonferroni correction for multiple testing of 11 haplotypes = 0.0045 (0.05/11). The differential effects on FHR-4 levels associated with each haplotype (elevated, baseline, or reduced) is indicated. b Box plot showing variations in log-transformed and centered FHR-4 levels with the four common AMD homozygous Chr1 diplotypes (Risk, Neutral, Prot-I62, and Prot-Del) (346 independent biological samples). c Comparison of Prot-I62 haplotypes with and without the effect allele (T) at rs7531555 generated using the IAMDGC cohort (13,378 controls and 17,541 cases) along with a box plot showing variations in log-transformed and centered FHR-4 levels among subjects stratified by rs800292 and rs7531555 diplotypes (149 independent biological samples). d Comparison of Risk haplotypes with and without the effect allele at rs10494745 and rs61818956 generated using the IAMDGC cohort (13,378 controls and 17,541 cases) along with a box plot showing variations in log-transformed and centered FHR-4 levels among subjects with Risk/Risk diplotypes stratified by rs10494745 (AA or GG) and rs61818956 (CC or TT) diplotypes (56 independent biological samples). e Schematic describing the effect of reduced or elevated FHR-4 levels on AMD susceptibility by haplotype group. Log-transformed OR and 95% CI were generated using the IAMDGC cohort (13,378 controls and 17,541 cases). In b, c, and d, OR, 95% CI and p-values (two-sided) for comparisons between haplotypes were obtained using logistic regressions including AMD case/control status as the dependent variable and age, sex, and the first two genetic principal components as covariates. In all box plots the horizontal center lines correspond to the medians of the log-transformed and centered FHR-4 distribution and the boxes delineate the 25th/75th percentile. The vertical solid lines represent the full range of the log-transformed and centered FHR-4 distribution in each group. Dots beyond this line indicate potential outliers. Associations with log-transformed and centered FHR-4 levels were assessed using the Kruskal–Wallis test. Post-hoc pairwise comparisons were performed using the Conover–Iman test. All p-values are two-sided and adjusted for multiple testing using Bonferroni corrections. Source data for these plots are provided as a Source Data file.