Fig. 3

Accuracy of mapping simulated paired-end reads to human population-specific augmented genome graphs. a The top principal components of a genomic relationship matrix constructed from autosomal variants detected in 2504 individuals that were included in phase 3 of the 1000 Genomes Project. The colored points indicate 405 samples from the GBR (European), YRI (African), STU (South Asia), and JPT (East Asia) populations. b Nucleotide diversity of the four populations calculated in non-overlapping 10 kb windows for variants of chromosome 19. The values below each boxplot indicate the nucleotide diversity for the four populations averaged across all sliding-windows. c Proportion of incorrectly mapped reads for four population-specific augmented genome graphs. d True-positive (sensitivity) and false-positive mapping rate (specificity) parameterized on mapping quality of the best performing graph from each population. e Read mapping accuracy for population-specific augmented graphs that contained variants that were either filtered for alternate allele frequency (triangles) or sampled randomly (circles) from all variants detected within a population. The dashed and solid line represents the average proportion of mapping errors across four populations using variant prioritization and random sampling, respectively. Results for single-end mapping are presented in Additional file 1: Fig. S6