Fig. 5

The workflow and performance of the ensemble variant genotyping method EVG. a The variant genotyping workflow of EVG mainly consists of three steps: (1) subsample sequencing reads, filter variants, and reformat the input variant VCF file; (2) select one or multiple suitable graph-based genotypers (shown as colored dots) and do genotyping with each of them in parallel; (3) merge the genotype results from step 2 and determine the final genotype for each variant. b Genotyping performance of SNPs, indels, ins & del (insertions and deletions), inversions, and duplications on simulated A. thaliana genomes under different sequencing depths (5×, 10×, 20×, 30×, 50×) and genome numbers (1, 15, 50). The genome graph for genotyping is constructed from the A. thaliana reference genome and different numbers of simulated alternative genomes. Paired-end short-reads (read length: 2 × 150 bp) are simulated for variant genotyping. For each genotyping scenario, the F-measure values of the other two best-performing genotypers are shown here. Transparent and solid bars represent the ability to predict variant “presence” (detection of variant regardless of the genotype) and exact “genotype” (requires both the detection of the variant and agreement between its called genotype and the true genotype). Detailed results are also provided in Additional file 2: Table S13