Haplotype Inference in Autotetraploids and Its Application to Genome-Wide Prediction and Association Analysis

For diploids, previous research has shown that haplotype-based approaches to genome-wide prediction and association analysis can improve upon single-marker methods by increasing statistical power and accuracy.   The objective of this research was to investigate whether similar benefits are observed in autotetraploids.  Two different haplotyping software packages—SATlotyper and polyHap—were evaluated using a simulated population of autotetraploids, which was created from five founders and 100 generations of random mating.  Haplotypes were inferred using 3­­–5 bi-allelic markers and compared with the frequency of true haplotypes from the simulation.  The accuracy of linkage disequilibrium estimates was also determined.   The effect of haplotype inference accuracy on statistical power for GWAS was investigated using simulated phenotypes.  Having validated the haplotype-based approach on simulated data, we compared its performance against single-marker analyses for GWAS and genome-wide prediction using empirical data from North American potato breeding programs.