Genomic selection shows improved expected genetic gain over phenotypic selection of agronomic traits in allotetraploid white clover

This article is a preprint and has not yet been submitted to a journal for publication.

Genomic selection, a valuable breeding tool to enhance genetic gain for low heritability traits, is applied widely in plants and animals, but only recently in forages. Our focus is white clover (Trifolium repens L.), an important legume of temperate agronomy due to its provision of quality feed and plant available nitrogen through symbiosis with soil-dwelling Rhizobium bacteria. Historically, this forage has shown little genetic improvement for dry matter yield (DMY) and persistence-related traits. Here we explore the feasibility of genomic selection in a white clover training population comprising 200 half-sibling (HS) families evaluated in a cattle-grazed field trial across three years and two locations. Combining these phenotype data and ~110,000 cost-effective genotyping-by-sequencing single nucleotide polymorphism markers among the HS-family parents, we assessed different types of two-stage genomic prediction models for DMY, growth score, leaf size and stolon traits. Predictive abilities were similar among the models and ranged from -0.17 to 0.44 across traits, and were stable for most traits when prediction model input was reduced to 100-120 HS-families and 5,500 markers. Incorporating a correlated trait in a multi-trait prediction model increased predictive ability by 28-124% compared with the primary trait alone. Based on deterministic modelling, integrating among HS-family phenotypic selection and within-family genomic selection at different selection pressures predicted DMY genetic gains 31% to 89% greater than HS-family phenotypic selection alone, despite a modest predictive ability of 0.3. This study confirms the potential economic advantages of integrating genomic selection with phenotypic selection to increase genetic gain significantly in white clover.