Linfeng Chen, Shouping Yang, Susan Araya, Charles Quigley, Earl Taliercio, Rouf Mian, James E. Specht, Brian W. Diers & Qijian Song

Theoretical and Applied Genetics May 2022; Vol. 135: 1797–1810

Key message

Software for high imputation accuracy in soybean was identified. Imputed dataset could significantly reduce the interval of genomic regions controlling traits, thus greatly improve the efficiency of candidate gene identification.

Abstract

Genotype imputation is a strategy to increase marker density of existing datasets without additional genotyping. We compared imputation performance of software BEAGLE 5.0, IMPUTE 5 and AlphaPlantImpute and tested software parameters that may help to improve imputation accuracy in soybean populations. Several factors including marker density, extent of linkage disequilibrium (LD), minor allele frequency (MAF), etc., were examined for their effects on imputation accuracy across different software. Our results showed that AlphaPlantImpute had a higher imputation accuracy than BEAGLE 5.0 or IMPUTE 5 tested in each soybean family, especially if the study progeny were genotyped with an extremely low number of markers. LD extent, MAF and reference panel size were positively correlated with imputation accuracy, a minimum number of 50 markers per chromosome and MAF of SNPs > 0.2 in soybean line were required to avoid a significant loss of imputation accuracy. Using the software, we imputed 5176 soybean lines in the soybean nested mapping population (NAM) with high-density markers of the 40 parents. The dataset containing 423,419 markers for 5176 lines and 40 parents was deposited at the Soybase. The imputed NAM dataset was further examined for the improvement of mapping quantitative trait loci (QTL) controlling soybean seed protein content. Most of the QTL identified were at identical or at similar position based on initial and imputed datasets; however, QTL intervals were greatly narrowed. The resulting genotypic dataset of NAM population will facilitate QTL mapping of traits and downstream applications. The information will also help to improve genotyping imputation accuracy in self-pollinated crops.

See https://link.springer.com/article/10.1007/s00122-022-04070-7

Fig. 2: Genotype imputation accuracy based on different numbers of SNPs in study panels. BEAGLE 5.0 and IMPUTE 5 were performed for population-based imputation, while AlphaPlantImpute was performed for bi-parental-based imputation. The imputation accuracy of filling and without-filling in AlphaPlantImpute were displayed