Next-generation sequencing (NGS) technologies enable high-throughput, low-cost genotyping in wheat. However, these methods generate large numbers of missing sites and several imputation methods have been developed to predict these missing data with accuracies dependent on reference selection, sequencing coverage, minor allele frequencies, etc. A graph-based computational framework called Practical Haplotype Graph (PHG) was recently introduced to efficiently store sequence based genotyping data and infer high-density genotypes by imputing low-coverage skim genotype sequences. The PHG serves as a database to store large-scale genomic variation in the form of pangenome haplotype and enables imputation of high-density genotypes from low-density genotyping platforms. In this study, we used whole exome capture sequencing dataset to develop a wheat PHG database. The objectives of this study were to estimate the accuracies of imputing whole exome capture genotypes from simulated skim-sequencing, genotyping-by-sequencing, and Illumina array data. We show that with a wheat PHG built to store WGS information for 95 diverse wheat genotypes, the PHG SNP-calling accuracy was minimally affected by sequencing coverage and imputation accuracies for low-coverage sequencing data ranged between 83% (0.01x) and 87% (1x). Beagle5.2 on the other hand could accurately impute low-coverage skim sequencing data with 81% accuracy (1x). The imputation with GBS was, surprisingly, higher with an average accuracy of 96%, and could prove to be very useful to achieve genomic selection goals at lower sequencing cost. It is yet to be determined how the genotypes imputed using PHG would help identify SNPs with a broader range of FHB resistance traits in genetic mapping experiments.