A reference panel for linkage disequilibrium and genotype imputation using whole-genome sequencing data from 2,680 participants across India

Figures

• 

  Figure 1. Linkage disequilibrium (LD) patterns in LASI-DAD compared with four super-populations from the 1000 Genomes Project (1000G) The four super-populations include African (AFR), European (EUR), South Asian (SAS), and East Asian (EAS). (A) Number of LD blocks identified by LDetect. (B) Distribution of LD block lengths in millions of base pairs (Mb), with LD blocks grouped into different length ranges for clear visualization. (C) Boxplots showing the distribution of average LD on chromosome 1, calculated between either all pairs of SNPs within each LD block or SNPs in each LD block and its adjacent two blocks. LD is evaluated as the squared correlation coefficient (r2) of genotypes between an SNP pair. (D) Distribution of variation in LD (varLD) scores evaluated between LASI-DAD and each of the four super-populations from 1000G. varLD scores were calculated as the sum of the absolute differences between the corresponding eigenvalues derived from the LD matrices of two comparing populations, using sliding windows of 50 SNPs each.

• 

  Figure 2. Linkage disequilibrium (LD) patterns in Indian sub-populations from LASI-DAD Individuals from LASI-DAD were grouped into three sub-populations based on their proportion of Ancestral North Indian ancestry (%ANI) and presence on the north/south cline of India to facilitate Indian sub-population analysis. The three sub-populations included the high %ANI group (0.540 ≤ %ANI < 0.824; n = 935), low %ANI group (%ANI ≤ 0.540; n = 941), and the out-of-the north/south cline group (n = 731). (A) Principal-component analysis (PCA) plot illustrating the genetic relationship among Indian, European (EUR), and East Asian (EAS) populations. PCA was conducted using Indian samples from LASI-DAD, along with EUR, SAS, and EAS samples from 1000G. The proportion of variance explained by each PC is indicated in parentheses. LASI-DAD individuals are colored by their geographic region of birth in India to highlight the relationship between population structure and geography in India. (B) The same PCA plot with LASI-DAD individuals colored by their sub-population groups. (C) Distribution of variation in LD (varLD) scores evaluated between LASI-DAD sub-populations. varLD scores were calculated as the sum of…

• 

  Figure 3. Number of autosomal variants included in different linkage disequilibrium (LD) reference panels Three LD reference panels were compared, including Indian samples from LASI-DAD (n = 2,607), South Asian (SAS) samples from 1000G (1000G SAS, n = 489), and SAS samples from TOPMed (TOP-LD SAS, n = 239). (A) Number of variants stratified by the range of minor allele frequency (MAF). (B) Number of variants with at least one other variant in LD under different r2 cutoffs, where r2 represents the squared Pearson correlation coefficient between the phased haplotypes of a variant pair. (C) Number of variant pairs stratified by different r2 cutoffs.

• 

  Figure 4. Polygenic risk score (PRS) analyses demonstrating the utility of LASI-DAD as an linkage disequilibrium (LD) reference panel (A) The predictive performance of PRS, as measured by incremental R2, was compared for PRS constructed with the LASI-DAD and 1000G South Asian (SAS) LD reference panels. Summary statistics for height and BMI were obtained from four studies of SAS ancestry, including a GWAS of height, an ExWAS of BMI from the GIANT consortium, and GWASs of both height and BMI from the Genes & Health study. PRS weights were estimated by PRS-CS. (B) The improvement in the predictive performance of PRS constructed with the LASI-DAD reference panel relative to the 1000G SAS reference panel. Error bars indicate 95% confidence intervals estimated from bootstrapping with 1,000 resamples.

• 

  Figure 5. Evaluating LASI-DAD as a reference panel for genotype imputation Compared reference panels include LASI-DAD, TOPMed, and GAsP. (A) Bar plots showing the number of genetic variants in LASI-DAD that can be imputed by each reference panel and by all three panels. (B) Imputation accuracy was evaluated for each reference panel, including meta-imputation to integrate the imputation results from either LASI-DAD and TOPMed, LASI-DAD and GAsP, or all three panels, across different MAF ranges. The MAF was calculated within the full 2,680 LASI-DAD samples. To ensure a fair comparison, the imputation accuracy was evaluated using the common set of variants that can be imputed by all three panels. Aggregated r2 was used as the evaluation metric, calculated by grouping variants within a specified MAF range, stacking their imputed dosages and genotype calls from the sequencing data into two separate vectors, and then computing the squared correlation coefficient between the two vectors.