Sparse testing designs for optimizing predictive ability in sugarcane populations

• DOI: 10.3389/fpls.2024.1400000
• Publication Date: 2024-07-23T05:12:55Z
• Authors: Julian Garcia-Aba...; Paul Adunola; Fernando Silva Ag...; Jhon Henry Trujil...; John Jaime Riascos; Reyna Persa; Julio Isidro y Sa...; Diego Jarquín
• Keywords: sparse testing designs; Optimization; 0301 basic medicine; Biología; Plant culture; Plant Science; genomic selection GS; SB1-1110; genomic prediction GP; 03 medical and health sciences; Genomic selection GS; Genomic prediction GP; Sugarcane breeding; Sparse testing designs; sugarcane breeding; optimization

Sugarcane is a crucial crop for sugar and bioenergy production. Saccharose content total weight are the two main key commercial traits that compose sugarcane’s yield. These under complex genetic control their response patterns influenced by genotype-by-environment (G×E) interaction. An efficient breeding of sugarcane demands an accurate assessment genotype stability through multi-environment trials (METs), where genotypes tested/evaluated across different environments. However, phenotyping all genotype-in-environment combinations often impractical due to cost limited availability propagation-materials. This study introduces sparse testing designs as viable alternative, leveraging genomic information predict unobserved prediction models. approach was applied dataset comprising 186 six environments (6×186=1,116 phenotypes). Our employed three predictive models, including environment, genotype, markers effects, well G×E saccharose accumulation (SA) tons cane per hectare (TCH). Calibration sets sizes varying between 72 (6.5%) (16.7%) number phenotypes were composed remaining 930 (83.3%). Additionally, we explored optimal common pattern prediction. Results demonstrate maximum accuracy SA ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" display="inline" id="im1"><mml:mrow><mml:mi>ρ</mml:mi><mml:mo>=</mml:mo><mml:mn>0.611</mml:mn></mml:mrow></mml:math> ) TCH id="im2"><mml:mrow><mml:mi>ρ=0.341</mml:mi></mml:mrow></mml:math> achieved using in training few (3) no (0) maximizing tested only once. Significantly, show reducing phenotypic records model calibration has minimal impact on ability, with 12 non-overlapped environment (72=12×6) being most convenient cost-benefit combination.