TABLE 1.

Predictive breeding (genomic prediction—GP, also known as genomic selection—GS) studies in forest tree species published during the last years.

Species	Populations	Trait data	Genotyping data	GP algorithm	Key conclusions	References
Elaeis guineensis	162 individuals from the Deli and Group B populations	Seven oil yield components	262 SSRs	PBLUP, GBLUP	Genomic selection (GBLUP) calibrated according to conditions of the experiment showed higher trait precision when using pedigree-based model	Cros et al., 2015
Elaeis guineensis	A × B hybrid progeny tests with almost 500 crosses for training and 200 crosses for independent validation	Seven oil yield components	(>5,000 GBS-derived SNPs	GBLUP, PBLUP	Preselection for yield components using GBS is the first possible application of GS in oil palm.	Cros et al., 2017
Hevea brasilensis	332 clones from the F1 cross PB 260 × RRIM 600	Rubber production	332 SSRs on site 1 and 296 SSRs on site 2	RKHS, BLR_A, RR-BLUP-A, BLR_AD, RR-BLUP_AD	Mean between-site GS accuracy reached 0.561 when using the 125–200 SSRs with the highest Ho. The simulations showed that by applying a genomic preselection among 3,000 seedlings in the nursery there is a greater precision of selection of the genomic preselection compared to the phenotypic preselection. Statistical method had no effect on GS precision	Cros et al., 2019
Eucalyptus grandis × E. urophylla hybrids	999 individuals from 45 families	Cellulose content, composition of lignin monomer, total lignin, WD	33,398 SNP	ABLUP, GBLUP, ssGBLUP	ssGBLUP is a tool with a great projection for the improvement of the precision and the bias of the classic GBLUP for the genomic evaluation in the improvement of Eucalyptus	Cappa et al., 2019
Picea abies	1,370 controlled-pollinated individuals from 46 unrelated parents	Quality features of solid wood, pilodyn penetration, acoustic speed	116,765 SNP	ABLUP-A, ABLUP-AD, GBLUP-AD, GBLUD-ADE	GBLUP-AD is a model with great utility in production and propagation. Tree breeders can use it for seedling selection, or family and full-siblings selection	Chen et al., 2019
Eucalyptus globulus	646 individuals out of approximately 10 individuals per family	WD, branch quality, DBH, HT	14,442 SNP	BRR, Bayes C, HAP, HAP-SNP	In general, the BRR and Bayes C methods had a higher predictive capacity for most of the traits. In particular, genomic models that included the haplotype effect (either HAP or HAP SNP) significantly increased the AP of traits with low heritability.	Ballesta et al., 2019
Eucalyptus cladocalyx	1,470 individuals from 49 families	DBH, HT, BHT, WD, STR, SLD, FI	3.8 K Illumina Infinium EUChip60K SNPs	Bayes A, Bayes B, Bayes C, BRR	An GSq approach outperformed GS models in terms of predictive ability when the proportion of the variance explained by the significant marker-trait associations was higher than those explained by the polygenic background and non-significant markers	Ballesta et al., 2020
Eucalyptus clones of E. urophylla× E. grandis	1,130 clones of 69 full- sib families	Biomass production, WUE, wood properties	3,303 SNPs	GBLUP	The inclusion of wood δ13C in the selection process may lead to Eucalyptus varieties adapted to marginal zones still presenting good performance for biomass and wood chemical traits	Bouvet et al., 2020
Picea abies	726 trees of 40 families of complete siblings from two localities	Density, microfiber angle, wood stiffness	5,660 Infinium iSelect SNP matrix SNPs from exome capture and sequencing	Single-trait: GBLUP, BRR, GBLUP, TGBLUP, ABLUP. Multi-traits: GBLUP	Genomic prediction models showed similar results, but the multi-trait model stood out when weevil attacks were not available. Most of the results indicate that the weevil resistance genotypes were higher when there was a greater proportion of height to diameter and greater rigidity of the wood.	Lenz et al., 2020
Pinus radiata	457 POP2 descendants of 63 parents, and 524 POP3 descendants of 24 parents	Branching frequency, stem straightness, internal verification, and external bleeding	1,371,123 exome sequencing capture SNPs	GBLUP, ABLUP	An efficient way to improve non-key traits is through genomic selection with a pedigree corrected using SNP information	Li et al., 2019
Pseudotsuga menziesii	13,615 individuals	HT, 13 environmental variables	66,969 SNPs	ssGBLUP	GS-PA can be substantially improved using ECs to explain environmental heterogeneity and G × E effects. The ssGBLUP methodology allows historical genetic trials containing non-genotyped samples to contribute in genomic prediction, and, thus, effectively boosting training population size which is a critical step	Ratcliffe et al., 2019
Shorea platyclados	356 individuals from a half-sib progeny population	Seven important traits, including growth, branching quality, wood quality traits	5,900 Illumina Hi-Seq X SNPs	rrBLUP	Selective breeding for these traits individually could be very effective, especially for increasing the diameter growth, branch diameter ratio and wood density simultaneously	Sawitri et al., 2020
Hevea brasiliensis	435 individual rubber trees at two sites. 252 F1 hybrids derived from a PR255 × PB217 cross, 146 F1 hybrids derived from a GT1 × RRIM701 cross, 37 genotypes from a GT1 × PB235 cross, and 4 testers (GT1, PB235, RRIM701, and RRIM600)	SC	30,546 GBS-derived SNPs	BLUP, SM, MM, MDs, Mde	Multi-environment models were superior to the single-environment genomic models. Methods in which GS is incorporated resulted in a fivefold increase in response to selection for SC with multi-environment GS (MM, MDe, or MDs)	Souza et al., 2019
Fraxinus excelsior	1,250 individuals	Tree health, ash dieback resistance	100–50,000 HiSeq X SNPs	RR-BLUP	Ash dieback resistance in F. excelsior is a polygenic trait that should respond well to both natural selection and breeding, which could be accelerated using genomic prediction	Stocks et al., 2019
Eucalyptus nitens	691 individuals	Solid wood production, height, DBH, stem straightness, WD, wood stiffness, wood shrinkage, growth strain	12,236 Illumina EUChip60K SNPs	BLUP, GBLUP	The greatest improvement in genetic parameters was obtained for tangential air-dry wood shrinkage and growth strain	Suontama et al., 2019
Pseudotsuga menziesii	A 38-year-old progeny test population (P1), selecting 37 of 165 families with complete siblings at random from 3 different settings. Validation population contained 247 descendants with controlled crosses from the 37 families	HT	Complete genotyping of exome capture	RR-BLUP, GRR, Byes-B	The validation of cross genomic selection of juvenile height in Douglas fir gave very similar results with the ABLUP predictive precision, but this precision may be linked to the relationship between training and validation conjugates	Thistlethwaite et al., 2019a
Pseudotsuga menziesii, Picea glauca, P. engelmannii	1,321 Douglas-fir trees, representing 37 full-sib F1 families and 1,126 interior spruce trees, representing 25 open-pollinated (half-sib) families	Mid-rotation height, WD	200–50,000 Illumina HiSeq 2000 SNPs	RR-BLUP	Reducing marker density cannot be recommended for carrying out GS in conifers. Significant LD between markers and putative causal variants was not detected using 50,000 SNPs	Thistlethwaite et al., 2020
Pinus contorta	Half- and full- sibs represented by 57 base parents and 42 full-sib families with an calculated effective population size of 92	Growth and wood quality	51,213 Illumina HiSeq SNPs	Bayes C, Bayes B, BLUP, GBLUP, ABLUP	The predictions of Marker-based models had accuracies that were equal to or better than pedigree-based models (ABLUP) when using several cross-validation scenarios and were better at ranking trees within families	Ukrainetz and Mansfield, 2020
Castanea dentate	7,173 descendants of BC3F3 from 346 “Clapper” mothers and 198 “Serious” mothers. For the BC3F2 progeny, a total of 1,134 “Clapper” and 1,042 “Graves” were sampled	Cryphonectria parasitica fungus severity (BC3F3) or presence/absence data (BC3F2)	Sequencing of a C. dentata clone in the PacBio Sequel platform	HBLUP, ABLUP, Bayes C	By means of genomic prediction and estimation of hybrid indices, a trade-off is between resistance and a proportion of inherited genome. The results found show that the genetic architecture underlying the heritability of resistance to blight is complex	Westbrook et al., 2020
Picea abies	484 progeny trees from 62 half-sib families	WD, MOE, MFA	130,269 Illumina HiSeq 2500 SNPs	ABLUP, GBLUP, rrBLUP, BayesB, RKHS	This study indicates standing tree-based measurements is a cost-effective alternative method for GS. Selection for density could be conducted at an earlier age than for MFA and MOE	Zhou et al. (2020)

For a comprehensive summary of previous studies not included here see Grattapaglia et al. (2018). Detailed abbreviations are shown at the end of the table. WUE, water use efficiency; SC, stem circumference; WD, wood density; MOE, modulus of elasticity; MFA, microfibril angle; DBH, diameter at breast height; HT, total tree height; BHT, first bifurcation height; STR, stem straightness; SLD, slenderness index; FI, flowering intensity; SNP, single nucleotide polymorphism; SSR, simple sequence repeat; GBS, genotyping by sequencing.