Genomic Prediction and Linkage Disequilibrium in Eucalyptus benthamii and Eucalyptus pellita using a 60K SNPs Chip (EUChip60K)

Eucalyptus benthamii (BEN) and Eucalyptus pellita (PEL) are species of increasing commercial interest due to their cold and drought tolerance, respectively. Genomic Selection (GS) is a promising approach to accelerate the breeding cycle and develop elite trees of these unexploited species. We report the development of genomic prediction models for growth traits and estimates of Linkage Disequilibrium (LD) decay in two breeding populations of BEN (Ne=53, n=505) and PEL (Ne=35, n=732), using SNPs genotyped with an Ilumina Infinium chip. Prediction models were initially built with 13,787 and 19,506 high quality polymorphic SNPs. Using progressively smaller SNP datasets (10,000 to 100), the results suggests that models with ~5,000 SNPs are equivalent in predictive ability (r_gy) to the full model for all traits and species. The r_gy inferred by Bayesian methods (BRR, BayesB, BayesA, BayesCπ, BL) reached similar estimates, varying from 0.14 for volume in BEN to 0.44 for DBH in PEL. The lower values of r_gy for BEN may be explained by the restricted occurrence of this species in its natural range and the limited genetic diversity sampled by this breeding population. The average genome-wide LD (r²) dropped below 0.2 within 15.6 and 70.6 Kbp for BEN and PEL, respectively. After correcting for relatedness and structure (r²VS) LD decayed faster, at 7.7 and 25.5 Kbp, confirming the strong effect of genetic relationship expected in these populations. This study sets the stage for the application of high-density SNPs and GS in two specialty eucalypts, to characterize their populations and advance breeding.