A multi-ancestry genetic reference for the Quebec population

Abstract

While international efforts have characterized genetic variation in millions of individuals, the interplay of environmental, social, cultural and genetic factors is poorly understood for most worldwide populations. The province of Quebec in Canada has been the site of numerous genetic studies, often focusing on Mendelian diseases in founder sub-populations. Here, we analyze genome-wide genotyped variation in 29,337 Quebec residents from the multi-ancestry population-based cohort CARTaGENE (CaG) who provided DNA samples. We also sequence the whole-genome of 2,173 CaG participants with four grandparents born in Canada (n = 1879), Haiti (n = 163) and Morocco (n = 131). We use this genetic information to gain insight into Quebec's demography and to help interpret the potential significance of variants identified in clinically important genes (e.g., SPG7 implicated in hereditary spastic paraplegia). We validate an imputation panel constructed by phasing the CaG whole-genome sequence data and find, using genome-wide association studies (GWAS) of 42 clinically relevant traits, that it increases the number of associated loci by ~7% when compared to results obtained after imputation with the larger TOPMed imputation panel. We provide allele frequency information and GWAS results through dedicated and publicly available websites. The genetic data, paired with phenotypic and environmental information, is available for global research use.

Fig. 1. Genetic structure in the CaG population-based cohort of Quebec.

A Principal components 1 and 2, labelled and colored by HDBSCAN clusters, as described in Methods. Word clouds of the countries of birth are provided for each cluster. Global population structure within the cohort is shaped by a combination of domestic structure and recent migration. B UMAP of the top 10 principal components of CaG, labelled and colored by the same HDBSCAN clusters as in A. Clusters reflect recent migration history, admixture, and regional genetic structure within Quebec. Some clusters, such as 10 and 12, are not evident in the 2D UMAP but form in the higher dimensional UMAP used in HDBSCAN. Word clouds for other variables are available in Supplementary Fig. 3. C, D Estimation of admixture proportions among CaG participants whose four grandparents were born in Morocco (QMO, C) or Haiti (QHA). ADMIXTURE estimation was run independently for each group for varied number K of components (Methods). Reported Ks were chosen by cross-validation analysis (Supplementary Figs. 6, 7 for all Ks). The plots show CaG individuals with four grandparents born in the same country, and three additional populations from the 1000 Genomes Project as reference in (D). YR Yoruba in Nigeria, CHB Han Chinese in Beijing, China; ACB African Caribbean in Barbados, Congo Dem. Rep. Democratic Republic of the Congo.

Fig. 2. Analysis of rare pathogenic variants in SPG7.

Structure of the SPG7 gene, highlighting its main three domains and the location of nine variants found in the CaG cohort. For each variant, amino acid/nucleotide change, allele count (AC), frequency in Quebec residents of French-Canadian ancestry (QFC), Morocco-born grandparents (QMO) and estimated mutation age (TMRCA) are displayed. The heatmaps show the carrier frequency of three SPG7 variants computed with ISGen in 24 historical regions of Quebec. The scale is given in number of carriers per 1000 individuals. Each variant has a different frequency distribution across Quebec. Highest frequency regions for each variant are highlighted: C.988-1 G > A (Saguenay and Charlevoix), p.Gly349Ser (Côte-de-Beaupré and Portneuf) and p.Asp765Asn (Bas-Saint-Laurent and Côte-du-Sud). Maps in Fig. 2 were created using the geopandas package in Python 3.12 (https://zenodo.org/records/3946761).

Fig. 3. Genome-wide statistically significant loci identified using different genotype imputation approaches in individuals of European genetic ancestry.

Each statistically significant independent locus was defined iteratively as the ±500-kb region centered on the variant with the lowest significant P-value, referred to as the lead variant. Any overlapping loci for the same trait were merged into a single locus, keeping one lead variant with the lowest P-value. Thus, by definition, each locus has only one lead variant. (A) The overlap between genome-wide statistically significant loci using three imputation approaches. Letters A, B, C, D, E, and F label the corresponding loci subsets. The number below the label shows the number of loci within each subset. The number of lead variants not found in the alternative reference panel is shown in parenthesis. B The X-axis shows the median of paired differences between imputation qualities at lead variants in CaG and TOPMed imputation results. The Y axis shows the median alternate allele frequency (AF) of lead variants based on CaG imputation results. The horizontal and vertical error bars show 95% confidence intervals calculated using 1,000,000 sets of variants: each set had the same number of variants as in the subsets in panel A, and variants in the simulated sets were matched on AF in the dataset imputed using the CaG imputation panel. Only statistically significant (significance threshold 0.025) two-tailed bootstrap P-values for the median of paired differences in imputation qualities are displayed below the subset labels to reduce image cluttering. C Each panel shows a fold change in the proportion of lead variants inside different subsets of significant loci, from left to right: lead variants for which alternate allele frequency (AF) in QFC individuals exceeds AF in gnomAD NFE, lead variants for which AF in QFC individuals is 2 times higher compared to gnomAD NFE, lead variants for which AF in QFC individuals is 4 times higher compared to gnomAD NFE. The P-values were computed as the proportion of 1,000,000 permuted samples that exceeded the observed fold change with a statistical significance threshold of 0.05.