Genomic relationship-aware sparse multi-environment trial design for optimizing breeding plot allocation

Phenotyping capacity remains a major constraint on the effective use of genomic selection in plant breeding. Sparse multi-environment trials (METs) can expand candidate evaluation under fixed resources but create a design trade-off: plots assigned to replication, overlap, or checks cannot also be used to observe new genotype-environment combinations. We developed GRACE-MET (Genomic Relationship-Aware Cross-Environment design for Multi-Environment Trials), a pre-phenotyping framework that treats sparse MET layout as a genomic resource-allocation problem. GRACE-MET uses marker-derived genomic relationships, multi-environment GBLUP reliability criteria, and field-layout constraints to rank candidate trial designs before planting. We evaluated GRACE-MET using empirical wheat, maize Genomes to Fields, and pea yield benchmarks, together with 50-repeat AlphaSimR simulations spanning family structure, genotype-by-environment interaction, overlap rate, partial replication, and field-layout error. Across empirical datasets, the best allocation strategy depended on environment transferability and the strength of non-genomic baselines. Low- or no-overlap designs were often competitive in wheat and maize, whereas the low-transfer pea benchmark favored broader local genotype-environment coverage over increased overlap. In known-truth simulations, family- and GRM-informed sparse designs improved missing-cell accuracy over non-genomic incomplete-block allocation in 33 of 36 settings, and family/GRM-informed P-Rep improved over conventional P-Rep in nearly all tested settings. However, P-Rep did not consistently outperform non-replicated sparse baselines because replicated plots displaced unique observations. These results show that sparse MET design is context-dependent. GRACE-MET provides a quantitative framework for comparing coverage, genomic connectedness, local precision, and environment-specific transferability before field resources are committed.