KBeagle: An Adaptive Strategy and Tool for Improving Imputation Accuracy and Computation Time

Whole-genome sequencing (WGS) technology has made significant progress in obtaining the genomic information of organisms and is now the primary way to uncover genetic variation. However, due to the complexity of the genome and technical limitations, large genome segments remain ungenotyped. Imputation is a useful strategy for predicting missing genotypes. The accuracy and computing speed of imputation software are important criteria that should inform future developments in genomic research. In this study, the K-Means algorithm and multithreading were used to cluster reference individuals to reduce the number and improve the length of haplotypes in the subpopulation. We named this strategy "KBeagle". In the comparison test, we determined that the KBeagle-imputed dataset (KID) can identify more single-nucleotide polymorphism (SNP) loci associated with the specified traits compared to the Beagle-imputed dataset (BID), while also achieving much lower false discovery rates (FDRs) and Type I error rates under the same power of detection of association signals. We envision that the main application of KBeagle will focus on livestock sequencing studies under a strong genetic structure. In summary, we have generated an accurate and efficient imputation method, improving the imputation matching rate and calculation time.