Fig 1.
Schematic overview of the HBimpute pipeline.
The values in brackets indicate the share of missing values in each step for the maize data set with 0.5X sequencing depths.
Table 1.
Discordance rates between the imputed sequence data and the 600k array data depending on the used imputation pipeline.
* For cells with a genotype call in STITCH itself, discordance rates were only 0.39% compared to 0.44 / 0.39% for HB-seq / HB-array for the same entries.
Fig 2.
Discordance rates of the imputed sequence data to the 600k array data depending on the used imputation pipeline and the allele frequency of the given variant.
Table 2.
Discordance rates between the imputed sequence data and high read-depth sequence data depending on the used imputation pipeline.
* For cells with a genotype call in STITCH itself, discordance rates were only 0.59% compared to 0.78 / 0.68% for HB-seq / HB-array for the same entries.
Fig 3.
Allele frequency spectrum of the genomic datasets for all bivarite markers that are shared between the array and sequence data panels.
Fig 4.
Estimated standardized read-depth for line PE0213 via the use of high read-depth sequence data (A/B), imputed low read-depth sequence data via HBimpute (C/D) and raw low depth depth sequence data (E/F) for chromosome 10 (A/C/E) and an exemplary chosen segment in a peak region (B/D/F).
Table 3.
Average predictive ability for nine maize traits [48] depending on the genotype data used for prediction.
The panel of overlapping markers includes all markers included in the array and sequence data panel after quality control filtering.
Fig 5.
Number of positive GWAS hits for simulated traits with 10 underlying QTL depending on the share of true positive hits (A). Median distance of the local GWAS peak (highest p-value) and the underlying true QTL for correct GWAS hits (B).
Fig 6.
Toy example for the HBimpute step.
Each column represents a SNP and each row represents a haplotype (for inbred lines: individual). Haplotype blocks are indicated by colored blocks. The blue and red block are overlapping.