Flow chart of this study.

Demographic characteristics of the SCOURGE Latin-American cohort.

A) Manhattan plot for the admixed AMR GWAS meta-analysis. Probability thresholds at p=5×10-8 and p=5×10-5 are indicated by the horizontal lines. Genome-wide significant associations with COVID-19 hospitalizations were found in chromosome 2 (within BAZ2B), chromosome 3 (within LZTFL1), chromosome 6 (within FOXP4), and chromosome 11 (within DDIAS). A Quantile-Quantile plot is shown in supplementary Figure 2. B) Regional association plots for rs1003835 at chromosome 2 and rs77599934 at chromosome 11; C) Allele frequency distribution across The 1000 Genomes Project populations for the lead variants rs1003835 and rs77599934.

Lead independent variants in the admixed AMR GWAS meta-analysis.

Novel variants in the SC-HGIALL and SC-HGI3POP meta-analyses (with respect to HGIv7). Independent signals after LD clumping.

Forest plot showing effect sizes and the corresponding confidence intervals for the sentinel variants identified in the AMR meta-analysis across populations. All beta values with their corresponding CIs were retrieved from the B2 population-specific meta-analysis from the HGI v7 release, except for AMR, for which the beta value and IC from the HGIAMR-SCOURGE meta-analysis is represented.

(A) Polygenic risk stratified by PGS deciles comparing each risk group against the lowest risk group (OR-95%CI); (B) Distribution of the PGS scores in each of the severity scale classes (0-Asymptomatic, 1-Mild disease, 2-Moderate disease, 3-Severe disease, 4-Critical disease).

Global Genetic Inferred Ancestry (GIA) composition in the SCOURGE Latin-American cohort.

European (EUR), African (AFR) and Native American (AMR) GIA was derived with ADMIXTURE from a reference panel composed of Aymaran, Mayan, Nahuan, and Quechuan individuals of Native-American genetic ancestry and randomly selected samples from the EUR and AFR 1KGP populations. The colours represent the different geographical sampling regions from which the admixed American individuals from SCOURGE were recruited.

Quantile-Quantile plot for the AMR GWAS meta-analysis.

A lambda inflation factor of 1.015 was obtained.

Regional association plots for the fine mapped loci in chromosomes 2 (upper panel) and 16 (lower panel).

Coloured in red, the variants allocated to the credible set at the 95% confidence according to the Bayesian fine mapping. In blue, the sentinel variant.

Sensitivity plots from COLOC with expression data from GTEx v8.

The range of p12 values (probability that a SNP is associated with both traits) for which the rule H4>0.7 is supported is shown in green in the right plots for each analysis. Plots in the left represent the variants included in the risk region common to both traits along their individual association -log10(p-values) for each trait, whereas the shading shows the posterior probability that the SNP is causal given H4 is true. Trait 1 corresponds to COVID-19 hospitalization, while trait 2 corresponds to gene expression in each analysis.

Sensitivity plots from COLOC with whole blood expression data from the GALA and SAGE II studies in AMR individuals.

AFRhp5 corresponds to the expression dataset computed in individuals with high African ancestries; AMRhp5 corresponds to the expression dataset computed individuals with high AMR ancestries; pooled corresponds to the dataset computed with the total of individuals from the study. In the right, the plots show in green the range of p12 values (probability that a SNP is associated with both traits) for which the rule H4>0.7 is supported. Plots in the left represent the variants included in the risk region common to both traits along their individual association -log10(p-values) for each trait, whereas the shading shows the posterior probability that the SNP is causal given H4 is true. Trait 1 corresponds to COVID-19 hospitalization, while trait 2 corresponds to gene expression.

Gene-tissue pairs for which either rs1003835 or rs60606421 are significant eQTLs at FDR<0.05 (data retrieved from https://gtexportal.org/home/snp/).

rs1003835 (chromosome 2) maps to BAZ2B, LY75, and PLA2R genes. As for the lead variant of chromosome 11, rs77599934, since it was not an eQTL, we used an LD proxy variant (rs60606421). DDIAS and PRCP genes map closely to this variant. NES and p-values correspond to the normalized effect size (and direction) of eQTL-gene associations and the p-value for the tissue, respectively.