Multi-omics prediction of terpene constituents and phenolic traits in Eucalyptus globulus using Bayesian models and tree-based machine learning

Terpenes and phenolic compounds are multifunctional plant metabolites that contribute to defense, signaling, and ecological interactions, while supporting industrial applications. Here, we implemented an integrative multi-omics framework to dissect the genetic and metabolic architecture of biobased compound production in a Eucalyptus globulus breeding population, a tree species of global economic importance. The dataset comprised 14,442 high-confidence SNPs genotyped using the EUChip60K array and 3,279 haplotype blocks, complemented by phenomic datasets derived from near-infrared (NIR) spectral absorbance (~ 900–2,500 nm) and pigment-related indices (chlorophyll, anthocyanins, flavonols, and nitrogen balance index). Gas chromatography-mass spectrometry (GC–MS) and ultra-high-performance liquid chromatography-quadrupole-time-of-flight tandem MS (UHPLC–QqTOF–MS2) revealed substantial inter-individual variation: essential oils were dominated by 1,8-cineole (42.70% ± 7.19%) and α-pinene (38.02% ± 5.20%), while phenolic extracts were enriched in phloroglucinol derivatives (92.67% ± 1.59%), notably the macrocarpal C isomer (12.98% ± 1.48%). Total additive heritability integrating SNP- and pedigree-based effects was high for MeOH extract yield (h2 = 0.721), total phenolics (h2 = 0.646), α-terpineol (h2 = 0.718), and 1,8-cineole (h2 = 0.658), indicating strong genetic control. Predictive modeling revealed that haplotype-based Bayesian approaches outperformed SNP- and phenomic-based models, with the highest accuracies for α-terpineol (prediction accuracy = 0.717), α-pinene (prediction accuracy = 0.554), and methanolic extract yield (prediction accuracy = 0.624). Feature selection identified pleiotropic markers (e.g., SNP12778, HAP205, HAP272, NIR2429) co-localized with genes involved in photosynthesis, signaling, and metabolic regulation. These findings lay the foundation for omics-assisted E. globulus breeding aimed at generating elite genotypes for large-scale production of high-value bioactive compounds with practical applications in health, agriculture, and bioenergy.